{"id":717840,"date":"2025-05-03T11:40:14","date_gmt":"2025-05-03T18:40:14","guid":{"rendered":"https:\/\/chinadigitaltimes.net\/chinese\/?p=717840"},"modified":"2025-05-03T15:01:40","modified_gmt":"2025-05-03T22:01:40","slug":"%e3%80%90404%e6%96%87%e5%ba%93%e3%80%91%e5%8d%8f%e5%92%8c%e5%8c%bb%e5%ad%a6%e9%99%a2%e8%a7%84%e5%9f%b9%e5%8c%bb%e7%94%9f%e8%91%a3%e8%a2%ad%e8%8e%b9%e5%8d%9a%e5%a3%ab%e8%ae%ba%e6%96%87%ef%bc%8c","status":"publish","type":"post","link":"https:\/\/chinadigitaltimes.net\/chinese\/717840.html","title":{"rendered":"\u3010404\u6587\u5e93\u3011\u534f\u548c\u533b\u5b66\u9662\u89c4\u57f9\u533b\u751f\u8463\u88ad\u83b9\u535a\u58eb\u8bba\u6587\uff0c \u4e2d\u56fd\u77e5\u7f51\u5df2\u5220\u9664"},"content":{"rendered":"<p><strong>CDT\u7f16\u8f91\u6ce8\uff1a<\/strong>\u65e5\u524d\uff0c\u4e2d\u65e5\u53cb\u597d\u533b\u9662\u533b\u751f\u8096\u98de\u88ab\u5176\u59bb\u4e3e\u62a5\u5a5a\u5916\u60c5\uff0c\u5176\u4e2d\u4e00\u4f4d\u5f53\u4e8b\u4eba\u4e3a\u89c4\u57f9\u533b\u5e08\u3001\u534f\u548c\u533b\u5b66\u9662\u535a\u58eb\u751f\u8463\u88ad\u83b9\u3002\u4e3e\u62a5\u4fe1\u4e2d\u8fd8\u6307\u51fa\uff0c\u8096\u98de\u66fe\u5728\u624b\u672f\u8fc7\u7a0b\u4e2d\u4e0e\u62a4\u58eb\u53d1\u751f\u51b2\u7a81\uff0c\u671f\u95f4\u4e0d\u987e\u60a3\u8005\u5b89\u5371\uff0c\u4e0e\u8463\u88ad\u83b9\u4e00\u540c\u79bb\u5f00\u624b\u672f\u5ba4\u3002\u968f\u540e\uff0c\u8463\u88ad\u83b9\u7684\u6559\u80b2\u80cc\u666f\u5f15\u53d1\u7f51\u7edc\u5173\u6ce8\u3002\u516c\u5f00\u4fe1\u606f\u663e\u793a\uff0c\u8463\u88ad\u83b9\u672c\u79d1\u5c31\u8bfb\u4e8e\u7f8e\u56fd\u54e5\u4f26\u6bd4\u4e9a\u5927\u5b66\u4e0b\u5c5e\u7684\u5973\u5b50\u5b66\u9662\u2014\u2014\u5df4\u7eb3\u5fb7\u5b66\u9662\uff0c\u4e3b\u4fee\u7ecf\u6d4e\u5b66\uff0c2019\u5e74\u88ab\u5317\u4eac\u534f\u548c\u533b\u5b66\u9662\u201c4+4\u201d\u4e34\u5e8a\u533b\u5b66\u957f\u5b66\u5236\u8bd5\u70b9\u73ed\u5f55\u53d6\u4e3a<a href=\"https:\/\/chinadigitaltimes.net\/chinese\/717874.html\" title=\"\u535a\u58eb\u7814\u7a76\u751f\">\u535a\u58eb\u7814\u7a76\u751f<\/a>\u3002\u76f8\u6bd4\u4e2d\u56fd\u4f20\u7edf\u533b\u5b66\u6559\u80b2\u4f53\u7cfb\u901a\u5e38\u66f4\u957f\u7684\u6559\u5b66\u4e0e\u89c4\u57f9\u5468\u671f\uff0c\u534f\u548c\u201c4+4\u201d\u6a21\u5f0f\u5728\u516c\u5e73\u6027\u4e0e\u4e13\u4e1a\u6027\u65b9\u9762<a href=\"https:\/\/chinadigitaltimes.net\/chinese\/717813.html\" title=\"\u5f15\u53d1\u8d28\u7591\">\u5f15\u53d1\u8d28\u7591<\/a>\u3002\u4e0e\u6b64\u540c\u65f6\uff0c\u6709\u7f51\u6c11\u6307\u51fa\uff0c\u8463\u88ad\u83b9\u7684\u535a\u58eb\u8bba\u6587\u4e0e\u5317\u4eac\u79d1\u6280\u5927\u5b66\u7684\u4e00\u9879\u53d1\u660e\u4e13\u5229\u5b58\u5728<a href=\"https:\/\/archive.is\/DEOet\" title=\"\u591a\u5904\u96f7\u540c\" target=\"_blank\" rel=\"noopener\">\u591a\u5904\u96f7\u540c<\/a>\uff0c\u6d89\u5acc\u5b66\u672f\u4e0d\u7aef\u3002\u4e8b\u4ef6\u6301\u7eed\u53d1\u9175\u540e\uff0c\u4e2d\u56fd\u77e5\u7f51\u5df2\u5c06\u5176\u535a\u58eb\u8bba\u6587\u4e0b\u67b6\uff0c\u800c\u8fd9\u7bc7\u8bba\u6587\u7684\u4e0b\u67b6\u6d41\u7a0b\u662f\u5426\u9075\u5faa\u65e2\u5b9a\u7684<a href=\"https:\/\/www.cjstp.cn\/CN\/10.11946\/cjstp.202302070063\" title=\"\u64a4\u7a3f\u6807\u51c6\" target=\"_blank\" rel=\"noopener\">\u64a4\u7a3f\u6807\u51c6<\/a>\u4ea6\u53d7\u5230\u8d28\u7591\u3002\u6709\u7f51\u53cb\u5728\u6b64\u4e4b\u524d\u4fdd\u5b58\u4e86\u8bba\u6587PDF\uff0c\u5e76\u4e0a\u4f20\u81f3GitHub\u3002\u4ee5\u4e0b\u6587\u5b57\u548c\u56fe\u7247\u7531CDT\u901a\u8fc7PDF\u7248\u8f6c\u5f55\u5b58\u6863\u3002<\/p>\n<p><strong>\u5317\u4eac\u534f\u548c\u533b\u5b66\u9662\u4e34\u5e8a\u533b\u5b66\u4e13\u4e1a\u6bd5\u4e1a\u8bba\u6587<\/strong><br \/>\n<strong>\u5b66\u6821\u4ee3\u7801\uff1a10023<\/strong><br \/>\n<strong>\u5b66 \u53f7\uff1a B2019012012<\/strong><\/p>\n<p><strong>\u8de8\u6a21\u6001\u56fe\u50cf\u878d\u5408\u6280\u672f\u5728\u533b\u7597\u5f71\u50cf\u5206\u6790\u4e2d\u7684\u7814\u7a76<\/strong><\/p>\n<p><strong>\u4e13\u4e1a\u5e74\u7ea7\uff1a\u5317\u4eac\u534f\u548c\u533b\u5b66\u9662\u4e34\u5e8a\u533b\u5b66\u4e13\u4e1a 2019 \u7ea7\u8bd5\u70b9\u73ed<\/strong><br \/>\n<strong>\u59d3\u540d\uff1a \u8463\u88ad\u83b9<\/strong><br \/>\n<strong>\u5bfc\u5e08\uff1a \u90b1\u8d35\u5174\uff08\u6559\u6388\uff09<\/strong><br \/>\n<strong>\u5317\u4eac\u534f\u548c\u533b\u5b66\u9662 \u4e34\u5e8a\u5b66\u9662\uff08\u5317\u4eac\u534f\u548c\u533b\u9662\uff09<\/strong><br \/>\n<strong>\u9aa8\u79d1<\/strong><br \/>\n<strong>\u5b8c\u6210\u65e5\u671f\uff1a2023\u5e745\u6708<\/strong><\/p>\n<p><strong>\u76ee\u5f55<\/strong><\/p>\n<p>\u6458\u8981 &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230; 1<br \/>\nAbstract &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;. 3<\/p>\n<ol>\n<li>\u57fa\u4e8e\u7279\u5f81\u5339\u914d\u7684\u8de8\u6a21\u6001\u56fe\u50cf\u878d\u5408\u7684\u5bab\u9888\u764c\u75c5\u53d8\u533a\u57df\u68c0\u6d4b &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230; 6<br \/>\n1.1. \u524d\u8a00 &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;.. 6<br \/>\n1.2. \u7814\u7a76\u65b9\u6cd5 &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230; 7<br \/>\n1.2.1. \u7814\u7a76\u8bbe\u8ba1\u548c\u5de5\u4f5c\u6d41\u7a0b &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;.. 7<br \/>\n1.2.2. \u8de8\u6a21\u6001\u56fe\u50cf\u878d\u5408 &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;. 7<br \/>\n1.2.3. \u5bab\u9888\u764c\u75c5\u53d8\u533a\u57df\u68c0\u6d4b &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230; 11<br \/>\n1.3. \u5b9e\u9a8c &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230; 11<br \/>\n1.3.1. \u4e34\u5e8a\u4fe1\u606f\u548c\u5f71\u50cf\u6570\u636e\u96c6 &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;.. 11<br \/>\n1.3.2. \u6a21\u578b\u8bad\u7ec3\u8fc7\u7a0b &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230; 12<br \/>\n1.3.3. \u8bc4\u4ef7\u6307\u6807 &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;.. 13<br \/>\n1.3.4. \u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u7684\u7ed3\u679c\u4e0e\u5206\u6790 &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230; 14<br \/>\n1.4. \u8ba8\u8bba &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230; 15<br \/>\n1.5. \u7ed3\u8bba &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230; 16<\/li>\n<li>\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u6570\u636e\u878d\u5408\u7684\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u7684\u8bca\u65ad &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;. 18<br \/>\n2.1. \u524d\u8a00 &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230; 18<br \/>\n2.2. \u7814\u7a76\u65b9\u6cd5 &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;. 19<br \/>\n2.2.1. \u7814\u7a76\u8bbe\u8ba1\u548c\u5de5\u4f5c\u6d41\u7a0b &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230; 19<br \/>\n2.2.2. \u9aa8\u8f6c\u79fb\u76ee\u6807\u533a\u57df\u68c0\u6d4b &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230; 20<br \/>\n2.2.3. \u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u6570\u636e\u878d\u5408 &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;. 20<br \/>\n2.2.4. \u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u7684\u5206\u7c7b\u6a21\u578b &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;. 21<br \/>\n2.3. \u5b9e\u9a8c &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230; 22<br \/>\n2.3.1. \u4e34\u5e8a\u4fe1\u606f\u548c\u5f71\u50cf\u6570\u636e\u96c6 &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;.. 22<br \/>\n2.3.2. \u6a21\u578b\u8bad\u7ec3\u8fc7\u7a0b &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230; 23<br \/>\n2.3.3. \u8bc4\u4ef7\u6307\u6807 &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;.. 24<br \/>\n2.3.4. \u5355\u6a21\u6001\u9aa8\u8f6c\u79fb\u7076\u68c0\u6d4b\u6a21\u578b\u53ca\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u5206\u7c7b\u6a21\u578b\u7684\u7ed3\u679c\u4e0e\u5206\u6790 &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;.. 25<br \/>\n2.4. \u8ba8\u8bba &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230; 30<br \/>\n2.5. \u7ed3\u8bba &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230; 32<br \/>\n\u5168\u6587\u5c0f\u7ed3 &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;.. 33<br \/>\n\u53c2\u8003\u6587\u732e &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;.. 35<br \/>\n\u7f29\u7565\u8bcd\u8868 &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;.. 40<br \/>\n\u6587\u732e\u7efc\u8ff0 &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;.. 41<\/li>\n<li>\u8de8\u6a21\u6001\u6df1\u5ea6\u5b66\u4e60\u6280\u672f\u5728\u4e34\u5e8a\u5f71\u50cf\u4e2d\u7684\u5e94\u7528 &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;. 41<br \/>\n3.1 Preface &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230; 41<br \/>\n3.2. Deep Neural Network (DNN) &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;.. 42<br \/>\n3.2.1. Supervised learning &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;. 43<br \/>\n3.2.2. Backpropagation &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230; 46<br \/>\n3.2.3. Convolutional neural networks (CNN) &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;. 46<br \/>\n3.3. Cross-modal fusion &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;. 49<br \/>\n3.3.1. Cross-modal fusion methods &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;.. 50<br \/>\n3.3.2. Cross-modal image translation &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;.. 51<br \/>\n3.4. The application of cross-modal deep learning &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;.. 52<br \/>\n3.5. conclusion &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230; 55<br \/>\n\u53c2\u8003\u6587\u732e &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230; 57<br \/>\n\u81f4\u8c22 &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;. 60<br \/>\n\u72ec\u521b\u6027\u58f0\u660e &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;. 61<br \/>\n\u5b66\u4f4d\u8bba\u6587\u7248\u6743\u4f7f\u7528\u6388\u6743\u4e66 &#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;&#8230;. 61<\/li>\n<\/ol>\n<p><strong>\u6458\u8981<\/strong><\/p>\n<p><strong>\u80cc\u666f<\/strong><\/p>\n<p>\u5f71\u50cf\u5b66\u68c0\u67e5\u662f\u533b\u7597\u9886\u57df\u6700\u5e38\u7528\u7684\u7b5b\u67e5\u624b\u6bb5\uff0c\u636e\u7edf\u8ba1\uff0c\u533b\u7597\u6570\u636e\u603b\u91cf\u4e2d\u6709\u8d85\u8fc790%\u662f\u7531\u5f71\u50cf\u6570\u636e\u6784\u6210[1]\u3002\u7136\u800c\uff0c\u6839\u636e\u4eb2\u8eab\u53c2\u4e0e\u7684\u4e34\u5e8a\u75c5\u4f8b[2]\u53ef\u77e5,\u5f88\u591a\u60c5\u51b5\u4e0b\uff0c\u4ec5\u51ed\u533b\u751f\u7684\u8089\u773c\u89c2\u5bdf\u548c\u4e3b\u89c2\u8bca\u65ad\u7ecf\u9a8c\uff0c\u4e0d\u8db3\u4ee5\u5bf9\u5f71\u50cf\u5b66\u5f02\u5e38\u4f5c\u4e00\u660e\u786e\u5224\u65ad\u3002\u800c\u8bca\u65ad\u4e0d\u660e\u5f15\u8d77\u7684\u9891\u7e41\u5c31\u533b\u3001\u8d3b\u8bef\u75c5\u60c5\uff0c\u5219\u4f1a\u4e25\u91cd\u5f71\u54cd\u60a3\u8005\u7684\u751f\u6d3b\u8d28\u91cf\u3002<\/p>\n<p>\u76f8\u8f83\u4e8e\u4f20\u7edf\u7684\u4e3b\u89c2\u9605\u7247\uff0c\u4eba\u5de5\u667a\u80fd\u6280\u672f\u901a\u8fc7\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u5206\u6790\u5927\u91cf\u5f71\u50cf\u548c\u8bca\u65ad\u6570\u636e\uff0c\u5b66\u4e60\u5bf9\u75c5\u7406\u8bca\u65ad\u6709\u7528\u7684\u7279\u5f81\uff0c\u5728\u5ba2\u89c2\u6570\u636e\u7684\u652f\u6301\u4e0b\u505a\u51fa\u66f4\u51c6\u786e\u7684\u5224\u65ad\u3002\u4e3a\u4e86\u6a21\u62df\u4e34\u5e8a\u533b\u751f\u7ed3\u5408\u5404\u79cd\u6210\u50cf\u6a21\u5f0f\uff08\u5982 CT\u3001MRI \u548c PET\uff09\u5f62\u6210\u8bca\u65ad\u7684\u8fc7\u7a0b\uff0c\u672c\u9879\u76ee\u91c7\u7528\u8de8\u6a21\u6001\u6df1\u5ea6\u5b66\u4e60\u65b9\u6cd5\uff0c\u5c06\u5404\u79cd\u5f71\u50cf\u5b66\u6a21\u6001\u7279\u5f81\u8fdb\u884c\u6709\u673a\u7ed3\u5408\uff0c\u5145\u5206\u5229\u7528\u5176\u5404\u81ea\u7684\u72ec\u7279\u4f18\u52bf\u8bad\u7ec3\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\uff0c\u4ee5\u63d0\u9ad8\u6a21\u578b\u6027\u80fd\u3002\u9274\u4e8e\u80bf\u7624\u76f8\u5173\u7684\u5f71\u50cf\u5b66\u8d44\u6599\u76f8\u5bf9\u4e30\u5bcc\uff0c\u672c\u9879\u76ee\u4ee5\u5bab\u9888\u764c\u548c\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u4e3a\u4f8b\uff0c\u6d4b\u8bd5\u4e86\u8de8\u6a21\u6001\u6df1\u5ea6\u5b66\u4e60\u65b9\u6cd5\u5728\u75c5\u53d8\u533a\u57df\u5b9a\u4f4d\u548c\u8f85\u52a9\u8bca\u65ad\u65b9\u9762\u7684\u6027\u80fd\uff0c\u4ee5\u89e3\u51b3\u4e34\u5e8a\u5b9e\u9645\u95ee\u9898\u3002<\/p>\n<p><strong>\u65b9\u6cd5<\/strong><\/p>\n<p>\u7b2c\u4e00\u90e8\u5206\u56de\u987e\u6027\u7eb3\u5165\u4e86220 \u4f8b\u6709FDG-PET\/CT \u6570\u636e\u7684\u5bab\u9888\u764c\u60a3\u8005\uff0c\u5171\u8ba172,602\u5f20\u5207\u7247\u56fe\u50cf\u3002\u5e94\u7528\u591a\u79cd\u56fe\u50cf\u9884\u5904\u7406\u7b56\u7565\u5bf9PET \u548cCT \u56fe\u50cf\u8fdb\u884c\u56fe\u50cf\u589e\u5f3a\uff0c\u5e76\u8fdb\u884c\u611f\u5174\u8da3\u533a\u57df\u8fb9\u7f18\u68c0\u6d4b\u3001\u81ea\u9002\u5e94\u5b9a\u4f4d\u548c\u8de8\u6a21\u6001\u56fe\u50cf\u5bf9\u9f50\u3002\u5c06\u5bf9\u9f50\u540e\u7684\u56fe\u50cf\u5728\u901a\u9053\u4e0a\u7ea7\u8054\u8f93\u5165\u76ee\u6807\u68c0\u6d4b\u7f51\u7edc\u8fdb\u884c\u68c0\u6d4b\u3001\u5206\u6790\u53ca\u7ed3\u679c\u8bc4\u4f30\u3002\u901a\u8fc7\u4e0e\u4f7f\u7528\u5355\u4e00\u6a21\u6001\u56fe\u50cf\u53ca\u5176\u4ed6 PET-CT \u878d\u5408\u65b9\u6cd5\u8fdb\u884c\u6bd4\u8f83\uff0c\u9a8c\u8bc1\u672c\u9879\u76ee\u63d0\u51fa\u7684 PET-CT \u81ea\u9002\u5e94\u533a\u57df\u7279\u5f81\u878d\u5408\u7ed3\u679c\u5728\u63d0\u9ad8\u6a21\u578b\u76ee\u6807\u68c0\u6d4b\u6027\u80fd\u65b9\u9762\u5177\u6709\u663e\u8457\u6027\u4f18\u52bf\u3002\u7b2c\u4e8c\u90e8\u5206\u56de\u987e\u6027\u7eb3\u5165\u4e86233 \u4f8b\u4e73\u817a\u764c\u60a3\u8005\uff0c\u6bcf\u4f8b\u6837\u672c\u5305\u542b CT\u3001MRI\u3001\u6216 PET \u4e00\u81f3\u4e09\u79cd\u6a21\u6001\u7684\u5168\u8eab\u5f71\u50cf\u6570\u636e\uff0c\u5171\u67093051 \u5f20CT \u5207\u7247\uff0c3543 \u5f20MRI \u5207\u7247\uff0c1818 \u5f20PET \u5207\u7247\u3002\u9996\u5148\u8bad\u7ec3YOLOv5 \u5bf9\u6bcf\u79cd\u5355\u4e00\u6a21\u6001\u56fe\u50cf\u4e2d\u7684\u9aa8\u8f6c\u79fb\u75c5\u7076\u8fdb\u884c\u76ee\u6807\u68c0\u6d4b\u3002\u6839\u636e\u68c0\u6d4b\u6846\u7684\u7f6e\u4fe1\u5ea6\u5212\u5206\u516b\u4e2a\u533a\u95f4\uff0c\u7edf\u8ba1\u6bcf\u4e2a\u5f71\u50cf\u5e8f\u5217\u4e0d\u540c\u7f6e\u4fe1\u5ea6\u533a\u95f4\u4e2d\u542b\u6709\u68c0\u51fa\u9aa8\u8f6c\u79fb\u75c5\u7076\u7684\u4e2a\u6570\uff0c\u5e76\u4ee5\u6b64\u5f52\u4e00\u5316\u540e\u4f5c\u4e3a\u7ed3\u6784\u5316\u533b\u7597\u7279\u5f81\u6570\u636e\uff0c\u91c7\u7528\u7ea7\u8054\u65b9\u5f0f\u878d\u5408\u4e09\u79cd\u6a21\u6001\u7684\u7ed3\u6784\u5316\u7279\u5f81\u5b9e\u73b0\u8de8\u6a21\u6001\u7279\u5f81\u878d\u5408\u3002\u518d\u7528\u591a\u79cd\u5206\u7c7b\u6a21\u578b\u5bf9\u7ed3\u6784\u5316\u6570\u636e\u8fdb\u884c\u5206\u7c7b\u548c\u8bc4\u4f30\u3002\u5c06\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u878d\u5408\u6570\u636e\u4e0e\u7279\u5f81\u8f6c\u6362\u540e\u7684\u5355\u6a21\u6001\u7ed3\u6784\u5316\u6570\u636e\uff0c\u4ee5\u53ca\u57fa\u4e8e C3D \u5206\u7c7b\u6a21\u578b\u7684\u524d\u878d\u5408\u65b9\u5f0f\u8fdb\u884c\u6bd4\u8f83\uff0c\u9a8c\u8bc1\u7b2c\u4e8c\u90e8\u5206\u63d0\u51fa\u7684\u65b9\u6cd5\u5728\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u8bca\u65ad\u4efb\u52a1\u4e2d\u7684\u4f18\u8d8a\u6027\u80fd\u3002<\/p>\n<p><strong>\u7ed3\u679c<\/strong><\/p>\n<p>\u7b2c\u4e00\u90e8\u5206\u7684\u57fa\u4e8e\u8de8\u6a21\u6001\u878d\u5408\u7684\u80bf\u7624\u68c0\u6d4b\u5b9e\u9a8c\u8bc1\u660e\uff0cPET-CT \u81ea\u9002\u5e94\u533a\u57df\u7279\u5f81\u878d\u5408\u56fe\u50cf\u663e\u8457\u63d0\u9ad8\u4e86\u5bab\u9888\u764c\u75c5\u53d8\u533a\u57df\u68c0\u6d4b\u7684\u51c6\u786e\u6027\u3002\u76f8\u6bd4\u4f7f\u7528CT \u6216PET \u5355\u6a21\u6001\u56fe\u50cf\u4ee5\u53ca\u5176\u4ed6\u878d\u5408\u65b9\u6cd5\u751f\u6210\u7684\u591a\u6a21\u6001\u56fe\u50cf\u4f5c\u4e3a\u7f51\u7edc\u8f93\u5165\uff0c\u76ee\u6807\u68c0\u6d4b\u7684\u5e73\u5747\u7cbe\u786e\u5ea6\u5206\u522b\u63d0\u9ad8\u4e86 6.06%\u548c 8.9%\uff0c\u4e14\u6d88\u9664\u4e86\u4e00\u4e9b\u5047\u9633\u6027\u7ed3\u679c\u3002\u4e0a\u8ff0\u6d4b\u8bd5\u7ed3\u679c\u5728\u4f7f\u7528\u4e0d\u540c\u7684\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u7684\u60c5\u51b5\u4e0b\u4fdd\u6301\u4e00\u81f4\uff0c\u8fd9\u8868\u660e\u81ea\u9002\u5e94\u8de8\u6a21\u6001\u878d\u5408\u65b9\u6cd5\u6709\u826f\u597d\u7684\u901a\u7528\u6027\uff0c\u53ef\u4ee5\u6cdb\u5316\u5e94\u7528\u4e8e\u5404\u79cd\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u7684\u9884\u5904\u7406\u9636\u6bb5\u3002\u7b2c\u4e8c\u90e8\u5206\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u75c5\u4f8b\u5206\u7c7b\u5b9e\u9a8c\u8bc1\u660e\uff0c\u8de8\u6a21\u6001\u878d\u5408\u6570\u636e\u663e\u8457\u63d0\u9ad8\u4e86\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u8bca\u65ad\u4efb\u52a1\u7684\u6027\u80fd\u3002\u76f8\u8f83\u4e8e\u5355\u6a21\u6001\u6570\u636e\uff0c\u8de8\u6a21\u6001\u878d\u5408\u6570\u636e\u7684\u5e73\u5747\u51c6\u786e\u7387\u548cAUC\u5206\u522b\u63d0\u9ad8\u4e867.9%\u548c8.5%\uff0c\u89c2\u5bdf ROC \u66f2\u7ebf\u548c PR \u66f2\u7ebf\u7684\u5f62\u72b6\u548c\u9762\u79ef\u4e5f\u5177\u6709\u76f8\u540c\u7684\u5b9e\u9a8c\u7ed3\u8bba\uff1a\u5728\u4e0d\u540c\u7684\u5206\u7c7b\u6a21\u578b\u4e2d\uff0c\u4f7f\u7528\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u6570\u636e\uff0c\u76f8\u6bd4\u5355\u6a21\u6001\u6570\u636e\uff0c\u5bf9\u4e8e\u9aa8\u8f6c\u79fb\u75c5\u4f8b\u7684\u5206\u7c7b\u6027\u80fd\u66f4\u4e3a\u4f18\u8d8a\u3002\u800c\u76f8\u8f83\u4e8e\u57fa\u4e8e C3D \u7684\u524d\u878d\u5408\u5206\u7c7b\u6a21\u578b\uff0c\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u540e\u878d\u5408\u7b56\u7565\u5728\u5206\u7c7b\u4efb\u52a1\u65b9\u9762\u7684\u6027\u80fd\u66f4\u4f18\u3002<\/p>\n<p><strong>\u7ed3\u8bba<\/strong><\/p>\n<p>\u672c\u9879\u76ee\u4e3b\u8981\u5305\u542b\u4e24\u4e2a\u90e8\u5206\u3002\u7b2c\u4e00\u90e8\u5206\u8bc1\u5b9e\u4e86\u57fa\u4e8e\u533a\u57df\u7279\u5f81\u5339\u914d\u7684\u8de8\u6a21\u6001\u56fe\u50cf\u878d\u5408\u540e\u7684\u6570\u636e\u96c6\u5728\u68c0\u6d4b\u6027\u80fd\u4e0a\u4f18\u4e8e\u5355\u6a21\u6001\u533b\u5b66\u56fe\u50cf\u6570\u636e\u96c6\u548c\u5176\u4ed6\u878d\u5408\u65b9\u6cd5\u3002\u7b2c\u4e8c\u90e8\u5206\u63d0\u51fa\u4e86\u4e00\u79cd\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u6570\u636e\u878d\u5408\u65b9\u6cd5\u3002\u4f7f\u7528\u878d\u5408\u540e\u7684\u6570\u636e\u8fdb\u884c\u5206\u7c7b\u4efb\u52a1\uff0c\u5176\u5206\u7c7b\u6027\u80fd\u4f18\u4e8e\u4ec5\u4f7f\u7528\u5355\u6a21\u6001\u6570\u636e\u8fdb\u884c\u5206\u7c7b\u6216\u4f7f\u7528\u524d\u878d\u5408\u65b9\u6cd5\u7684\u6027\u80fd\u3002\u6839\u636e\u4e0d\u540c\u6a21\u6001\u533b\u5b66\u56fe\u50cf\u7684\u7279\u5f81\u5dee\u5f02\u4e0e\u4e92\u8865\u6027\uff0c\u672c\u9879\u76ee\u9a8c\u8bc1\u4e86\u8de8\u6a21\u6001\u6df1\u5ea6\u5b66\u4e60\u6280\u672f\u5728\u75c5\u53d8\u533a\u57df\u5b9a\u4f4d\u548c\u8f85\u52a9\u8bca\u65ad\u65b9\u9762\u7684\u4f18\u52bf\u3002\u76f8\u6bd4\u4e8e\u53ea\u4f7f\u7528\u5355\u6a21\u6001\u6570\u636e\u8fdb\u884c\u8bad\u7ec3\u7684\u6a21\u578b\uff0c\u8de8\u6a21\u6001\u6df1\u5ea6\u5b66\u4e60\u6280\u672f\u6709\u66f4\u4f18\u7684\u8bca\u65ad\u51c6\u786e\u7387\uff0c\u53ef\u4ee5\u6709\u6548\u7684\u6210\u4e3a\u4e34\u5e8a\u8f85\u52a9\u5de5\u5177\uff0c\u534f\u52a9\u548c\u6307\u5bfc\u4e34\u5e8a\u51b3\u7b56\u3002<\/p>\n<p><strong>\u5173\u952e\u8bcd\uff1a\u8de8\u6a21\u6001\u878d\u5408\uff0c\u6df1\u5ea6\u5b66\u4e60\uff0c\u5f71\u50cf\u5206\u6790\uff0c\u5bab\u9888\u764c\uff0c\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb<\/strong><\/p>\n<p><strong>Abstract<\/strong><\/p>\n<p><strong>Background<\/strong><\/p>\n<p>Imaging examinations serve as the predominant screening method in the medical field. As statistics reveal, imaging data constitute over 90% of the entire medical dataset. Nonetheless, clinical cases have demonstrated that mere subjective diagnoses by clinicians often fall short in making definitive judgments on imaging anomalies. Misdiagnoses or undiagnosed conditions, which result in frequent hospital visits and delayed treatment, can profoundly affect patients&#8217; quality of life.<\/p>\n<p>Compared to the traditional subjective image interpretation by clinicians, AI leverages deep neural networks to analyze large-scale imaging and diagnostic data, extracting valuable features for pathology diagnosis, and thus facilitating more accurate decision-making, underpinned by objective data. To emulate clinicians&#8217; diagnostic process that integrates various imaging modalities like CT, MRI, and PET, a cross-modal deep learning methodology is employed. This approach synergistically merges features from different imaging modalities, capitalizing on their unique advantages to enhance model performance.<\/p>\n<p>Given the ample availability of oncologic imaging data, the project exemplifies the efficacy of this approach in cervical cancer segmentation and detection of breast cancer bone metastasis, thereby addressing pragmatic challenges in clinical practice.<\/p>\n<p><strong>Methods<\/strong><\/p>\n<p>The first part retrospectively analyzed 72,602 slices of FDG-PET\/CT scans from 220 cervical cancer patients. Various preprocessing strategies were applied to enhance PET and CT images, including edge detection, adaptive ROI localization, and cross-modal image<br \/>\nfusion. The fused images were then concatenated on a channel-wise basis and fed into the object detection network for the precise segmentation of cervical cancer lesions. Compared to single modality images (either CT or PET) and alternative PET-CT fusion techniques,<br \/>\nthe proposed method of PET-CT adaptive fusion was found to significantly enhance the object detection performance of the model. The second part of the study retrospectively analyzed 3,051 CT slices, 3,543 MRI slices and 1,818 PET slices from 233 breast cancer patients, with each case containing whole-body imaging of one to three modalities (CT, MRI, or PET). Initially, YOLOv5 was trained to detect bone metastases in images across different modalities. The confidence levels of the prediction boxes were segregated into eight tiers, following which the number of boxes predicting bone metastases in each imaging sequence was tallied within each confidence tier. This count was then normalized and utilized as a structured feature. The structured features from the three modalities were fused in a cascaded manner for cross-modal fusion. Subsequently, a variety of classification models were then employed to evaluate the structured features for diagnosing bone metastasis. In comparison to feature-transformed single-modal data and the C3D early fusion method, the cross-modal fusion data founded on feature transformation demonstrated superior performance in diagnosing breast cancer bone metastasis.<\/p>\n<p><strong>Results<\/strong><\/p>\n<p>The first part of our study delivered compelling experimental results, showing a significant improvement in the accuracy of cervical cancer segmentation when using adaptively fused PET-CT images. Our approach outperformed other object detection algorithms based on either single-modal images or multimodal images fused by other methods, with an average accuracy improvement of 6.06% and 8.9%, respectively, while also effectively mitigating false-positive results. These promising test results remained consistent across different object detection models, highlighting the robustness and universality of our adaptive fusion method, which can be generalized in the preprocessing stage of diverse object detection models. The second part of our study demonstrated that cross-modal fusion based on feature transformation could significantly improve the performance of bone metastasis classification models. When compared to algorithms employing single-modal data, models based on cross-modal data had an average increase in accuracy and AUC of 7.9% and 8.5%, respectively. This improvement was further corroborated by the shapes of the ROC and PR curves. Across a range of classification models, feature-transformed cross-modal data<br \/>\nconsistently outperformed single-modal data in diagnosing breast cancer bone metastasis. Moreover, late fusion strategies grounded in feature transformation exhibited superior performance in classification tasks when juxtaposed with early fusion methods such as C3D.<\/p>\n<p><strong>Conclusions<\/strong><\/p>\n<p>This project primarily consists of two parts. The first part substantiates that deep learning object detection networks founded on the adaptive cross-modal image fusion method outperform those based on single-modal images or alternative fusion methods. The second part presents a cross-modal fusion approach based on feature transformation. When the fused features are deployed for classification models, they outperform those utilizing solely single-modal data or the early fusion model. In light of the differences and complementarity in the features of various image modalities, this project underscores the strengths of cross-modal deep learning in lesion segmentation and disease classification. When compared to models trained only on single-modal data, cross-modal deep learning offers superior diagnostic accuracy, thereby serving as an effective tool to assist in clinical decision-making.<\/p>\n<p><strong>Keywords: cross-modal fusion, deep learning, image analysis, cervical cancer, breast<br \/>\ncancer bone metastasis<\/strong><\/p>\n<p><strong>1. \u57fa\u4e8e\u7279\u5f81\u5339\u914d\u7684\u8de8\u6a21\u6001\u56fe\u50cf\u878d\u5408\u7684\u5bab\u9888\u764c\u75c5\u53d8\u533a\u57df\u68c0\u6d4b<\/strong><\/p>\n<p><strong>1.1. \u524d\u8a00<\/strong><\/p>\n<p>\u5bab\u9888\u764c\u662f\u5973\u6027\u7fa4\u4f53\u4e2d\u53d1\u75c5\u7387\u7b2c\u56db\u4f4d\u7684\u764c\u75c7\uff0c\u6bcf\u5e74\u5f71\u54cd\u5168\u7403\u8fd1 50 \u4e07\u5973\u6027\u7684\u751f\u547d\u5065\u5eb7[3] \u3002\u51c6\u786e\u548c\u53ca\u65f6\u7684\u8bc6\u522b\u5bab\u9888\u764c\u81f3\u5173\u91cd\u8981\uff0c\u662f\u5426\u80fd\u5bf9\u5176\u8fdb\u884c\u65e9\u671f\u8bc6\u522b\u51b3\u5b9a\u4e86\u6cbb\u7597\u65b9\u6848\u7684\u9009\u62e9\u53ca\u9884\u540e\u60c5\u51b5[4]\u3002\u6c1f\u4ee3\u8131\u6c27\u8461\u8404\u7cd6\u6b63\u7535\u5b50\u53d1\u5c04\u8ba1\u7b97\u673a\u65ad\u5c42\u663e\u50cf\/\u7535\u5b50\u8ba1\u7b97\u673a\u65ad\u5c42\u626b\u63cf\uff08fluorodeoxyglucose-positron emission tomography\/computed tomography, FDG-PET\/CT\uff09\uff0c\u56e0\u5176\u4f18\u8d8a\u7684\u654f\u611f\u6027\u548c\u7279\u5f02\u6027\uff0c\u6210\u4e3a\u4e86\u4e00\u4e2a\u91cd\u8981\u7684\u5bab\u9888\u764c\u68c0\u6d4b\u65b9\u5f0f[5] \u3002\u7531\u4e8eCT \u80fd\u591f\u6e05\u6670\u5730\u663e\u793a\u89e3\u5256\u7ed3\u6784\uff0cFDG-PET \u80fd\u591f\u5f88\u597d\u5730\u53cd\u6620\u5c40\u7076\u7684\u4ee3\u8c22\u4fe1\u606f\u5f62\u6210\u529f\u80fd\u5f71\u50cf\uff0cFDG-PET\/CT \u878d\u5408\u56fe\u50cf\u5bf9\u53ef\u7591\u5bab\u9888\u764c\u75c5\u7076\u7684\u663e\u793a\u6bd4\u5355\u72ec\u4f7f\u7528\u9ad8\u5206\u8fa8\u7387 CT \u66f4\u51c6\u786e\uff0c\u7279\u522b\u662f\u5728\u68c0\u6d4b\u533a\u57df\u6dcb\u5df4\u7ed3\u53d7\u7d2f\u548c\u76c6\u8154\u5916\u75c5\u53d8\u6269\u5c55\u65b9\u9762[6] ,[7] ,[8] \u3002\u7136\u800c\uff0c\u7528\u4f20\u7edf\u65b9\u6cd5\u4e3a\u5355\u4e00\u60a3\u8005\u7684 FDG-PET\/CT \u6570\u636e\u8fdb\u884c\u5206\u6790\u9700\u8981\u9605\u8bfb\u6570\u767e\u5e45\u5f71\u50cf\uff0c\u5bf9\u75c5\u53d8\u533a\u57df\u8fdb\u884c\u9274\u522b\u5206\u6790\uff0c\u8fd9\u4e00\u6781\u4e3a\u8017\u65f6\u7684\u8fc7\u7a0b\u5df2\u7ecf\u59a8\u788d\u4e86\u4e34\u5e8a\u533b\u751f\u5bf9\u5b50\u5bab\u9888\u764c\u7684\u4e34\u5e8a\u8bca\u65ad\u3002<\/p>\n<p>\u968f\u7740\u8ba1\u7b97\u673a\u786c\u4ef6\u548c\u7b97\u6cd5\u7684\u8fdb\u6b65\uff0c\u5c24\u5176\u662f\u4ee5\u6df1\u5ea6\u5b66\u4e60[9] \u3001\u56fe\u50cf\u5904\u7406\u6280\u672f[10] ,[11] \u4e3a\u4ee3\u8868\u7684\u673a\u5668\u5b66\u4e60\u6280\u672f\u7684\u9769\u65b0\uff0c\u8fd9\u4e9b\u4eba\u5de5\u667a\u80fd\u7b97\u6cd5\u5728\u4e34\u5e8a\u533b\u5b66\u7684\u8bb8\u591a\u9886\u57df\u4e2d\u8d77\u7740\u5173\u952e\u4f5c\u7528[12]\u3002\u57fa\u4e8e\u5176\u5f3a\u5927\u6709\u6548\u7684\u7279\u5f81\u63d0\u53d6\u80fd\u529b[13] ,[14] \uff0c\u6df1\u5ea6\u5b66\u4e60\u4e2d\u7684\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\u53ef\u4ee5\u901a\u8fc7\u68af\u5ea6\u4e0b\u964d\u81ea\u52a8\u5b66\u4e60\u56fe\u50cf\u4e2d\u7684\u4e3b\u8981\u7279\u5f81[15 \uff0c\u6781\u5927\u5730\u63d0\u9ad8\u76ee\u6807\u8bc6\u522b\u7684\u51c6\u786e\u6027 [16]\uff0c\u4f7f\u6df1\u5ea6\u5b66\u4e60\u6210\u4e3a\u8ba1\u7b97\u673a\u56fe\u50cf\u5904\u7406\u9886\u57df\u7684\u4e3b\u6d41\u6280\u672f[17] ,[18]\u3002\u5229\u7528\u6df1\u5ea6\u5b66\u4e60\u6280\u672f\u5bf9\u5bab\u9888\u764c\u5f71\u50cf\u8fdb\u884c\u5206\u6790\u53ef\u4ee5\u8f85\u52a9\u4e34\u5e8a\u533b\u751f\u505a\u51fa\u66f4\u4e3a\u51c6\u786e\u7684\u5224\u65ad\uff0c\u51cf\u8f7b\u4e34\u5e8a\u533b\u751f\u7684\u5de5\u4f5c\u8d1f\u62c5\uff0c\u63d0\u9ad8\u8bca\u65ad\u7684\u51c6\u786e\u6027[19]\u3002<\/p>\n<p>\u76ee\u524d\u5df2\u7ecf\u6709\u5f88\u591a\u5728\u5355\u4e00\u6a21\u6001\u56fe\u50cf\u4e2d\uff08CT \u6216 PET\uff09\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u6280\u672f\u8fdb\u884c\u75c5\u53d8\u68c0\u6d4b\u7684\u5de5\u4f5c\uff1aSeung\u7b49\u4f7f\u7528\u673a\u5668\u5b66\u4e60\u6280\u672f\u4f9d\u636ePET\u56fe\u50cf\u5bf9\u9884\u6d4b\u80ba\u764c\u7ec4\u7ec7\u5b66\u4e9a\u578b[20]\uff1bSasank \u8fdb\u884c\u4e86\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u7b97\u6cd5\u68c0\u6d4b\u5934 CT \u4e2d\u5173\u952e\u4fe1\u606f\u7684\u56de\u987e\u6027\u7814\u7a76[21]\uff1bChen \u4f7f\u7528\u968f\u673a\u6e38\u8d70\uff08random walk\uff09\u548c\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u5bf9 CT \u56fe\u50cf\u4e2d\u7684\u80ba\u90e8\u75c5\u53d8\u8fdb\u884c\u5206\u5272[22] ,[23]\u3002\u4f46\u5f88\u5c11\u6709\u5173\u4e8e\u4f7f\u7528\u8de8\u6a21\u6001\u56fe\u50cf\u878d\u5408\u6df1\u5ea6\u5b66\u4e60\u65b9\u6cd5\u8fdb\u884c\u75c5\u53d8\u68c0\u6d4b\u7684\u7814\u7a76\u3002<\/p>\n<p>\u57fa\u4e8e PET\/CT \u878d\u5408\u56fe\u50cf\u7684\u75c5\u53d8\u68c0\u6d4b\u9879\u76ee\u5305\u62ec\u4e09\u4e2a\u7814\u7a76\u4efb\u52a1\uff1a\u533a\u57df\u7279\u5f81\u5339\u914d[24]\uff0c\u8de8\u6a21\u6001\u56fe\u50cf\u878d\u5408[25 \u548c\u76ee\u6807\u75c5\u53d8\u533a\u57df\u68c0\u6d4b[26]\u3002Mattes \u4f7f\u7528\u4e92\u4fe1\u606f\u4f5c\u4e3a\u76f8\u4f3c\u6027\u6807\u51c6\uff0c\u63d0\u51fa\u4e86\u4e00\u79cd\u4e09\u7ef4PET \u5411\u80f8\u90e8CT \u914d\u51c6\u7684\u533a\u57df\u7279\u5f81\u5339\u914d\u7b97\u6cd5[27]\u3002Maqsood \u63d0\u51fa\u4e86\u4e00\u79cd\u57fa\u4e8e\u53cc\u5c3a\u5ea6\u56fe\u50cf\u5206\u89e3\u548c\u7a00\u758f\u8868\u793a\u7684\u8de8\u6a21\u6001\u56fe\u50cf\u878d\u5408\u65b9\u6848[28]\u3002Elakkiya \u5229\u7528\u66f4\u5feb\u7684\u57fa\u4e8e\u533a\u57df\u7684\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\uff08Faster Region-Based Convolutional Neural Network, FR-CNN\uff09\u8fdb\u884c\u9888\u90e8\u6591\u70b9\u7684\u68c0\u6d4b[29]\u3002\u76ee\u524d\u8fd8\u6ca1\u6709\u5c06\u4e0a\u8ff0\u4e09\u4e2a\u7814\u7a76\u4efb\u52a1\uff0c\u5373\u533a\u57df\u7279\u5f81\u5339\u914d\u3001\u8de8\u6a21\u6001\u56fe\u50cf\u878d\u5408\u3001\u75c5\u53d8\u533a\u57df\u68c0\u6d4b\u4efb\u52a1\uff0c\u7ed3\u5408\u8d77\u6765\u7684\u7814\u7a76\u5de5\u4f5c\u3002<\/p>\n<p>\u4e3a\u4e86\u51cf\u8f7b\u4e34\u5e8a\u533b\u751f\u7684\u5de5\u4f5c\u8d1f\u62c5\uff0c\u57fa\u4e8e\u8de8\u6a21\u6001\u6df1\u5ea6\u5b66\u4e60\u65b9\u6cd5\uff0c\u672c\u9879\u76ee\u7684\u7b2c\u4e00\u90e8\u5206\u63d0\u51fa\u4e86\u4e00\u4e2a\u7edf\u4e00\u7684\u591a\u6a21\u6001\u56fe\u50cf\u878d\u5408\u548c\u76ee\u6807\u68c0\u6d4b\u6846\u67b6\uff0c\u7528\u4e8e\u5bab\u9888\u764c\u7684\u68c0\u6d4b\u3002<\/p>\n<p><strong>1.2. \u7814\u7a76\u65b9\u6cd5<\/strong><\/p>\n<p><strong>1.2.1. \u7814\u7a76\u8bbe\u8ba1\u548c\u5de5\u4f5c\u6d41\u7a0b<\/strong><\/p>\n<p>\u672c\u9879\u76ee\u65e8\u5728\u68c0\u6d4b CT \u548c PET \u56fe\u50cf\u4e2d\u5bab\u9888\u764c\u7684\u75c5\u53d8\u533a\u57df\uff0c\u5de5\u4f5c\u6d41\u7a0b\u5982\u56fe 1-1 \u6240\u793a\uff1a\u626b\u63cf\u8bbe\u5907\u5bf9\u6bcf\u4f4d\u60a3\u8005\u8fdb\u884cPET \u548cCT \u56fe\u50cf\u5e8f\u5217\u7684\u91c7\u96c6\uff1b\u901a\u8fc7\u533a\u57df\u7279\u5f81\u5339\u914d\u548c\u56fe\u50cf\u878d\u5408\u6765\u5408\u6210\u6e05\u6670\u4e14\u4fe1\u606f\u4e30\u5bcc\u7684\u8de8\u6a21\u6001\u56fe\u50cf\u878d\u5408\u7ed3\u679c\uff1b\u91c7\u7528\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u7684\u76ee\u6807\u68c0\u6d4b\u65b9\u6cd5\u5728\u878d\u5408\u56fe\u50cf\u4e2d\u5bf9\u53ef\u7591\u5bab\u9888\u764c\u7684\u75c5\u53d8\u533a\u57df\u8fdb\u884c\u76ee\u6807\u68c0\u6d4b\u3002\u5728\u56fe 1-1 \u7684\u6700\u540e\u4e00\u884c\u4e2d\uff0c\u77e9\u5f62\u6846\u51fa\u7684\u9ec4\u8272\u533a\u57df\u53ca\u56fe\u4e2d\u53f3\u4e0a\u89d2\u653e\u5927\u7684\u533a\u57df\u4e2d\u5c55\u793a\u4e86\u68c0\u6d4b\u51fa\u7684\u5bab\u9888\u764c\u75c5\u53d8\u533a\u57df\u3002<\/p>\n<p>\u56fe 1-1 \u5de5\u4f5c\u6d41\u7a0b<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-11.54.31\u202fAM.png\" alt=\"img\" \/><\/p>\n<p><strong>1.2.2. \u8de8\u6a21\u6001\u56fe\u50cf\u878d\u5408<\/strong><\/p>\n<p>\u56fe 1-2 \u5c55\u793a\u4e86\u8de8\u6a21\u6001\u56fe\u50cf\u878d\u5408\u7b97\u6cd5\u7684\u6d41\u7a0b\u56fe\u3002\u6839\u636e\u8ba1\u7b97\u53d1\u73b0\u4e24\u79cd\u6a21\u6001\u56fe\u50cf\u7684\u6bd4\u4f8b\u548c\u4f4d\u7f6e\u4e0d\u540c\uff0c\u5982\u4ec5\u8fdb\u884c\u7b80\u5355\u7684\u878d\u5408\u4f1a\u9519\u8bef\u5730\u5c06\u5904\u4e8e\u4e0d\u540c\u4f4d\u7f6e\u7684\u7ec4\u7ec7\u5f71\u50cf\u91cd\u53e0\uff0c\u4ece\u800c\u4f7f\u7ec4\u7ec7\u53d1\u751f\u9519\u4f4d\uff0c\u5b9a\u4f4d\u4e0d\u51c6\uff0c\u4ea7\u751f\u4e0d\u53ef\u63a5\u53d7\u7684\u8bef\u5dee\u3002\u56e0\u6b64\uff0c\u7b2c\u4e00\u90e8\u5206\u63d0\u51fa\u4e86\u4e00\u79cd\u8de8\u6a21\u6001\u56fe\u50cf\u878d\u5408\u7b56\u7565\uff0c\u5176\u4e2d\u7684\u6b65\u9aa4\u5305\u62ec\u5bf9\u611f\u5174\u8da3\u533a\u57df\uff08region of interest, ROI\uff09\u7684\u81ea\u9002\u5e94\u5b9a\u4f4d\u548c\u56fe\u50cf\u878d\u5408\u3002<\/p>\n<p>\u5728PET \u548cCT \u56fe\u50cf\u4e2d\uff0c\u81ea\u9002\u5e94ROI\u5b9a\u4f4d\u80fd\u591f\u7cbe\u51c6\u8bc6\u522b\u5f85\u5206\u6790\u5904\u7406\u7684\u5173\u952e\u76ee\u6807\uff0c\u5373\u4eba\u4f53\u7ec4\u7ec7\u5f71\u50cf\uff0c\u7136\u540e\u8ba1\u7b97\u4e0d\u540c\u6a21\u6001\u56fe\u50cf\u4e0b\u7ec4\u7ec7\u5f71\u50cf\u4e4b\u95f4\u7684\u6bd4\u4f8b\u548c\u4f4d\u79fb\u3002\u4f9d\u636e\u4e0a\u8ff0\u8ba1\u7b97\u7ed3\u679c\u901a\u8fc7\u7f29\u653e\u3001\u586b\u5145\u548c\u88c1\u526a\u7684\u65b9\u5f0f\u6765\u878d\u5408 PET \u548cCT \u56fe\u50cf\u3002<\/p>\n<p>\u56fe 1-2 CT \u548cPET \u8de8\u6a21\u6001\u56fe\u50cf\u878d\u5408\u7b97\u6cd5\u7684\u6d41\u7a0b\u56fe<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-11.56.26\u202fAM.png\" alt=\"img\" \/><\/p>\n<p><strong>1.2.2.1. \u81ea\u9002\u5e94ROI \u5b9a\u4f4d<\/strong><\/p>\n<p>\u9274\u4e8e\u6570\u636e\u96c6\u4e2d PET \u56fe\u50cf\u4e0e CT \u56fe\u50cf\u7684\u9ed1\u8272\u80cc\u666f\u5747\u4e3a\u96f6\u50cf\u7d20\u503c\u586b\u5145\uff0cROI \u5185\u975e\u96f6\u50cf\u7d20\u503c\u8f83\u591a\uff0c\u800c ROI \u8fb9\u7f18\u7684\u975e\u96f6\u50cf\u7d20\u503c\u8f83\u5c11\uff0c\u56e0\u6b64\uff0c\u9009\u7528\u7ebf\u68c0\u6d4b\u65b9\u6cd5\u6765\u6807\u753b\u4e24\u79cd\u6a21\u6001\u56fe\u50cf\u4e2d\u7684 ROI\uff0c\u6700\u7ec8\u6807\u5212\u7ed3\u679c\u5982\u56fe 1-2 \u4e2d\u7684\u7eff\u8272\u7ebf\u6846\u51fa\u7684\u90e8\u5206\u6240\u793a\uff0c\u8fd9\u56db\u6761\u7ebf\u662f ROI \u5728\u56db\u4e2a\u65b9\u5411\u4e0a\u7684\u8fb9\u754c\u3002\u5728\u4e0d\u540c\u65b9\u5411\u4e0a\u8ba1\u7b97\u6bd4\u4f8b\u5c3a\u3002\u5728\u5c06 PET \u56fe\u50cf\u653e\u5927\u540e\uff0c\u6839\u636eROI\u5b9e\u73b0CT \u548cPET \u56fe\u50cf\u7684\u50cf\u7d20\u7ea7\u5bf9\u9f50\u3002\u88c1\u526a\u6389\u591a\u4f59\u7684\u533a\u57df\uff0c\u5e76\u7528\u96f6\u50cf\u7d20\u503c\u6765\u8865\u5145\u7a7a\u767d\u533a\u57df\u3002\u5982\u56fe 1- 3\uff08a\uff09\u6240\u793a\uff0c\u7ebf\u68c0\u6d4b\u65b9\u6cd5\u4ece\u4e2d\u5fc3\u70b9\u51fa\u53d1\uff0c\u5411\u56db\u4e2a\u65b9\u5411\u5373\u4e0a\u4e0b\u5de6\u53f3\u5bf9\u975e\u96f6\u50cf\u7d20\u503c\u8fdb\u884c\u904d\u5386\uff0c\u5e76\u8bb0\u5f55\u4e0b\u884c\u6216\u5217\u4e0a\u7684\u975e\u96f6\u50cf\u7d20\u503c\u7684\u6570\u91cf\u3002\u5982\u56fe1- 3\uff08b\uff09\u6240\u793a\uff0c\u7ea2\u8272\u7bad\u5934\u4ee3\u8868\u904d\u5386\u7684\u65b9\u5411\u3002\u5728\u4ece ROI \u4e2d\u5fc3\u5411\u8fb9\u7f18\u8fdb\u884c\u904d\u5386\u65f6\uff0c\u6cbf\u904d\u5386\u7ecf\u7ebf\u4e0a\u7684\u975e\u96f6\u50cf\u7d20\u503c\u6570\u91cf\u9010\u6e10\u51cf\u5c11\uff0c\u5982\u679c\u67d0\u4e00\u7ebf\u4e0a\u975e\u96f6\u50cf\u7d20\u503c\u7684\u8ba1\u6570\u4f4e\u4e8e\u9884\u8bbe\u7684\u9608\u503c\uff0c\u90a3\u4e48\u610f\u5473\u7740\u8be5\u7ebf\u5df2\u7ecf\u89e6\u53ca\u5230 ROI \u7684\u8fb9\u7f18\uff0c\u5982\u56fe 1- 3\uff08c\uff09\u6240\u793a\u3002\u7136\u800c\uff0c\u5982\u679c\u76f4\u63a5\u5bf9\u672a\u7ecf\u9884\u5904\u7406\u7684\u56fe\u50cf\u5e94\u7528\u7ebf\u68c0\u6d4b\u65b9\u6cd5\uff0c\u4f1a\u56e0\u53d7\u6a21\u7cca\u8fb9\u7f18\u53ca\u5176\u566a\u58f0\u7684\u5f71\u54cd\uff0c\u5f97\u5230\u8f83\u5dee\u7684\u5bf9\u9f50\u7ed3\u679c\uff0c\u96be\u4ee5\u8bbe\u7f6e\u9608\u503c\u3002\u56e0\u6b64\uff0c\u9700\u5bf9PET \u548cCT \u56fe\u50cf\u5355\u72ec\u6267\u884c\u56fe\u50cf\u589e\u5f3a\u9884\u5904\u7406\uff0c\u4ee5\u4f18\u5316 ROI \u6807\u5316\u7ed3\u679c\uff0c\u6539\u5584\u8de8\u6a21\u6001\u878d\u5408\u6548\u679c\u3002<\/p>\n<p>\u7531\u4e8ePET \u548cCT \u56fe\u50cf\u5177\u6709\u4e0d\u540c\u7684\u7eb9\u7406\u7279\u5f81\uff0c\u5e94\u7528\u4e0d\u540c\u7684\u9884\u5904\u7406\u7b56\u7565\uff0c\u5206\u522b\u5bf9\u56fe\u50cf\u8fdb\u884c\u589e\u5f3a\u5904\u7406\uff0c\u4ee5\u5f3a\u5316 ROI \u7684\u8fb9\u7f18\u7279\u6027\uff0c\u540c\u65f6\u6d88\u9664\u566a\u58f0\u4ea7\u751f\u7684\u5e72\u6270\uff0c\u518d\u5728\u4e24\u79cd\u4e0d\u540c\u6a21\u6001\u56fe\u50cf\u4e2d\u8fdb\u884c ROI \u5b9a\u4f4d\uff0c\u5982\u56fe1-2 \u6240\u793a\u3002<\/p>\n<p><strong>\u56fe 1-3 ROI \u68c0\u6d4b\u793a\u610f\u56fe<\/strong><\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-11.57.30\u202fAM.png\" alt=\"img\" \/><\/p>\n<p>CT \u56fe\u50cf\u662f\u7528 X \u5c04\u7ebf\u5bf9\u68c0\u67e5\u90e8\u4f4d\u4e00\u5b9a\u539a\u5ea6\u7684\u7ec4\u7ec7\u5c42\u8fdb\u884c\u626b\u63cf\uff0c\u7531\u63a2\u6d4b\u5668\u63a5\u6536\u900f\u8fc7\u8be5\u5c42\u9762\u7684 X \u5c04\u7ebf\uff0c\u7ecf\u6570\u5b57\u8f6c\u6362\u5668\u53ca\u8ba1\u7b97\u673a\u5904\u7406\u540e\u5f62\u6210\u7684\u89e3\u5256\u5b66\u56fe\u50cf\u3002CT \u56fe\u50cf\u901a\u5e38\u6bd4 PET \u56fe\u50cf\u66f4\u6e05\u6670\u3002\u4e3a\u4e86\u63d0\u53d6 ROI\uff0c\u9700\u5229\u7528\u56fe\u50cf\u589e\u5f3a\u6280\u672f\u5bf9 CT \u56fe\u50cf\u8fdb\u884c\u9884\u5904\u7406\uff1a\u9996\u5148\uff0c\u901a\u8fc7\u56fe\u50cf\u9510\u5316\u589e\u5f3a\u8fb9\u7f18\u7279\u5f81\u548c\u7070\u5ea6\u8df3\u53d8\u90e8\u5206\uff0c\u4f7f CT \u56fe\u50cf\u7684\u8fb9\u7f18\uff08\u5373\u7070\u5ea6\u503c\u7a81\u53d8\u533a\u57df\uff09\u4fe1\u606f\u66f4\u52a0\u7a81\u51fa\uff1b\u7531\u4e8e\u9510\u5316\u53ef\u80fd\u5bfc\u81f4\u4e00\u5b9a\u7684\u566a\u58f0\uff0c\u518d\u4f7f\u7528\u9ad8\u65af\u6a21\u7cca\u6ee4\u6ce2\u5668\uff08Gaussian blur\uff09[30 \u8fdb\u884c\u56fe\u50cf\u5e73\u6ed1\u53bb\u566a\uff0c\u5c06\u566a\u58f0\u6240\u5728\u50cf\u7d20\u70b9\u5904\u7406\u4e3a\u5468\u56f4\u76f8\u90bb\u50cf\u7d20\u503c\u7684\u52a0\u6743\u5e73\u5747\u8fd1\u4f3c\u503c\uff0c\u6d88\u9664\u5f71\u54cd\u6210\u50cf\u8d28\u91cf\u7684\u8fb9\u7f18\u6bdb\u8e81\uff1b\u5e76\u6267\u884cCanny\u8fb9\u7f18\u68c0\u6d4b\uff08Canny edge detection \uff09[31 \u6765\u8bbe\u5b9a\u9608\u503c\u5e76\u8fde\u63a5\u8fb9\u7f18\uff0c\u4ece\u800c\u5728\u56fe\u50cf\u4e2d\u63d0\u53d6\u76ee\u6807\u5bf9\u8c61\u7684\u8fb9\u7f18\u3002\u5c3d\u7ba1Canny\u8fb9\u7f18\u68c0\u6d4b\u7b97\u6cd5\u5df2\u5305\u542b\u9ad8\u65af\u6a21\u7cca\u7684\u53bb\u566a\u64cd\u4f5c\uff0c\u4f46\u5b9e\u9a8c\u8bc1\u5b9e\u4e24\u6b21\u9ad8\u65af\u6a21\u7cca\u540e\u7684\u8fb9\u7f18\u63d0\u53d6\u6548\u679c\u66f4\u4f18\u3002\u5728\u5bf9\u56fe\u50cf\u8fdb\u884c\u9510\u5316\u5904\u7406\u540e\uff0c\u5c06\u63d0\u53d6\u7684\u8fb9\u7f18\u56fe\u50cf\u4e0e\u9ad8\u65af\u6a21\u7cca\u540e\u7684\u56fe\u50cf\u8fdb\u884c\u53e0\u52a0\u3002\u5177\u4f53\u5730\uff0c\u5bf9\u4e24\u4e2a\u56fe\u50cf\u4e2d\u7684\u6bcf\u4e2a\u50cf\u7d20\u76f4\u63a5\u8fdb\u884c\u50cf\u7d20\u503c\u76f8\u52a0\uff0c\u6700\u7ec8\u5f97\u5230\u8fb9\u7f18\u66f4\u6e05\u6670\u4e14\u51cf\u8f7b\u566a\u58f0\u5f71\u54cd\u7684\u589e\u5f3a\u540e CT \u6a21\u6001\u56fe\u50cf\u3002<\/p>\n<p>PET \u56fe\u50cf\u662f\u57fa\u4e8e\u95f4\u63a5\u63a2\u6d4b\u5230\u7684\u7531\u6b63\u7535\u5b50\u653e\u5c04\u6027\u6838\u7d20\u53d1\u5c04\u7684 \u03b3 \u5c04\u7ebf\uff0c\u7ecf\u8fc7\u8ba1\u7b97\u673a\u8fdb\u884c\u6563\u5c04\u548c\u968f\u673a\u4fe1\u606f\u7684\u6821\u6b63\uff0c\u5f62\u6210\u7684\u5f71\u50cf\uff0c\u80fd\u591f\u663e\u793a\u4f53\u5185\u4ee3\u8c22\u6d3b\u52a8\u7684\u4fe1\u606f\u3002\u5c3d\u7ba1PET \u53ef\u4ee5\u663e\u793a\u5206\u5b50\u4ee3\u8c22\u6c34\u5e73\uff0c\u4f46\u7531\u4e8e\u6210\u50cf\u539f\u7406\u7684\u5dee\u5f02\uff0cPET \u56fe\u50cf\u76f8\u8f83\u4e8e CT \u56fe\u50cf\u663e\u5f97\u6a21\u7cca\u3002\u5bf9PET \u7684\u9884\u5904\u7406\u65b9\u5f0f\u4e0e\u5bf9CT \u56fe\u50cf\u7684\u7c7b\u4f3c\uff0c\u4f46\u7701\u7565\u4e86\u9ad8\u65af\u6a21\u7cca\u5904\u7406\u56fe\u50cf\u566a\u58f0\u7684\u6b65\u9aa4\uff0c\u56e0\u4e3a\u5728\u9510\u5316 PET \u6a21\u6001\u56fe\u50cf\u540e\u4ea7\u751f\u7684\u566a\u58f0\u8f83\u5c11\uff0c\u4e3a\u9632\u6b62\u6709\u6548\u7279\u5f81\u4fe1\u606f\u7684\u4e22\u5931\uff0c\u7565\u8fc7\u8fd9\u4e00\u73af\u8282\u3002<\/p>\n<p>\u4e3a\u4e86\u5c06\u4e24\u4e2a\u6a21\u6001\u7684\u56fe\u50cf\u8fdb\u884c\u533a\u57df\u7279\u5f81\u5339\u914d\uff0c\u4f7f\u7528PET \u548cCT \u56fe\u50cf\u4e2d\u7684\u77e9\u5f62ROI\u6846\u6765\u8ba1\u7b97\u7f29\u653e\u6bd4\u4f8b\u548c\u4f4d\u79fb\u53c2\u6570\uff0c\u5e76\u901a\u8fc7\u7f29\u653e\u3001\u586b\u5145\u548c\u88c1\u526a\u64cd\u4f5c\u5bf9PET \u548cCT \u56fe\u50cf\u4e2d\u7684ROI \u8fdb\u884c\u5bf9\u9f50\u3002<\/p>\n<p><strong>1.2.2.2. \u56fe\u50cf\u878d\u5408<\/strong><\/p>\n<p>CT \u548c PET \u56fe\u50cf\u7684\u5c3a\u5bf8\u5206\u522b\u4e3a 512\u00d7512 \u50cf\u7d20\u548c 128\u00d7128 \u50cf\u7d20\uff0cROI \u7279\u5f81\u533a\u57df\u4f4d\u4e8e\u56fe\u50cf\u7684\u4e2d\u5fc3\u4f4d\u7f6e\u3002\u901a\u8fc7\u7f29\u653e\u3001\u96f6\u503c\u586b\u5145\u548c\u526a\u5207\uff0c\u653e\u5927PET \u56fe\u50cf\u7684\u5c3a\u5bf8\u4ee5\u4e0eCT \u56fe\u50cf\u7684\u5c3a\u5bf8\u4fdd\u6301\u4e00\u81f4\uff0c\u5e76\u4e14\u5c06\u4e24\u4e2a\u6a21\u6001\u56fe\u50cf\u4e4b\u95f4\u7684 ROI \u5bf9\u9f50\uff0c\u4ee5\u4fbf\u540e\u7eed\u7684\u878d\u5408\u5904\u7406\u3002\u7ecf\u5904\u7406\u7684PET \u548cCT \u56fe\u50cf\u8f6c\u5316\u4e3a\u7070\u5ea6\u5f62\u5f0f\uff0c\u5206\u522b\u8fdb\u884c\u52a0\u6743\u548c\u56fe\u50cf\u53e0\u52a0\uff0c\u5c06\u5176\u7f6e\u4e8e\u4e0d\u540c\u901a\u9053\u4e2d\uff0c\u4f5c\u4e3a\u7f51\u7edc\u7684\u8f93\u5165\u5c42\u3002\u7531\u4e8e PET \u56fe\u50cf\u80fd\u5c55\u793a\u4f53\u5185\u5206\u5b50\u5c42\u9762\u7684\u4ee3\u8c22\u6c34\u5e73\uff0c\u5176\u5bf9\u4e8e\u80bf\u7624\u68c0\u6d4b\u7684\u654f\u611f\u6027\u9ad8\u4e8eCT \u56fe\u50cf\u3002\u56e0\u6b64\uff0c\u672c\u7814\u7a76\u7684\u56fe\u50cf\u878d\u5408\u65b9\u6cd5\u4e3aPET \u56fe\u50cf\u7684ROI \u5206\u914d\u4e86\u66f4\u591a\u6743\u91cd\uff0c\u4ee5\u63d0\u9ad8\u5bab\u9888\u764c\u68c0\u6d4b\u4efb\u52a1\u7684\u8868\u73b0\u3002<\/p>\n<p>\u56fe 1-4 \u6bd4\u8f83\u4e86\u672c\u9879\u76ee\u63d0\u51fa\u7684\u81ea\u9002\u5e94\u56fe\u50cf\u878d\u5408\u7684\u7ed3\u679c\u548c\u76f4\u63a5\u878d\u5408\u7684\u7ed3\u679c\uff0c\u9009\u53d6\u4eba\u4f53\u4e0d\u540c\u90e8\u4f4d\u7684 CT\u3001PET \u548c PET\/CT \u56fe\u50cf\u7684\u878d\u5408\u7ed3\u679c\u8fdb\u884c\u5c55\u793a\u3002\u7b2c\u4e00\u548c\u7b2c\u4e8c\u5217\u5206\u522b\u5c55\u793a\u4e86\u672a\u7ecf\u5904\u7406\u7684\u539f\u59cb CT \u548c PET \u56fe\u50cf\u3002\u7b80\u5355\u878d\u5408\u7b97\u6cd5\u4ec5\u5c06\u4e24\u4e2a\u56fe\u50cf\u7684\u50cf\u7d20\u70b9\u76f8\u52a0\uff0c\u5e76\u672a\u6267\u884c\u7279\u5f81\u5339\u914d\u8fc7\u7a0b\uff0c\u5f97\u5230\u7684\u878d\u5408\u56fe\u50cf\u65e0\u4efb\u4f55\u5b9e\u7528\u4ef7\u503c\u3002\u7531\u4e8e\u901a\u9053\u62fc\u63a5\u878d\u5408\u540e\u7684\u56fe\u50cf\u8f6c\u53d8\u4e3a\u9ad8\u7ef4\u591a\u6a21\u6001\u6570\u636e\uff0c\u800c\u975e\u4e09\u901a\u9053\u6570\u5b57\u56fe\u50cf\uff0c\u56e0\u6b64\u56fe 1-4 \u5e76\u672a\u5c55\u793a\u901a\u9053\u62fc\u63a5\u878d\u5408\u65b9\u6cd5\u6240\u5f97\u56fe\u50cf\u3002\u800c\u672c\u9879\u76ee\u63d0\u51fa\u7684\u81ea\u9002\u5e94\u56fe\u50cf\u878d\u5408\u65b9\u6cd5\u5b9e\u73b0\u4e86\u8de8\u6a21\u6001\u56fe\u50cf\u7684\u7cbe\u51c6\u878d\u5408\uff0c\u53ef\u7528\u4e8e\u8fdb\u4e00\u6b65\u7684\u89c2\u5bdf\u548c\u8ba1\u7b97\u3002<\/p>\n<p>\u56fe 1-4 \u4e0d\u540c\u56fe\u50cf\u878d\u5408\u65b9\u5f0f\u7684\u53ef\u89c6\u5316\u7ed3\u679c<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-11.59.07\u202fAM.png\" alt=\"img\" \/><\/p>\n<p><strong>1.2.3. \u5bab\u9888\u764c\u75c5\u53d8\u533a\u57df\u68c0\u6d4b<\/strong><\/p>\n<p>\u5148\u7531\u4e24\u4f4d\u4e34\u5e8a\u533b\u751f\u5bf9\u8de8\u6a21\u6001\u878d\u5408\u56fe\u50cf\u4e2d\u7684\u75c5\u53d8\u533a\u57df\u8fdb\u884c\u4eba\u5de5\u6807\u6ce8\uff0c\u5e76\u8bad\u7ec3YOLOv5 [32] \u76ee\u6807\u68c0\u6d4b\u7f51\u7edc\u6765\u8bc6\u522b\u878d\u5408\u56fe\u50cf\u4e2d\u7684\u75c5\u7076\u533a\u57df\uff0c\u5982\u56fe 1-5 \u6240\u793a\u3002\u6a21\u5757\u9aa8\u67b6\u7528\u4e8e\u63d0\u53d6\u56fe\u50cf\u7684\u6df1\u5c42\u7279\u5f81\uff0c\u4e3a\u51cf\u5c11\u901a\u8fc7\u5207\u7247\u64cd\u4f5c\u8fdb\u884c\u91c7\u6837\u8fc7\u7a0b\u4e2d\u7684\u4fe1\u606f\u635f\u5931\uff0c\u91c7\u7528\u805a\u7126\u7ed3\u6784\uff0c\u5e76\u4f7f\u7528\u8de8\u9636\u6bb5\u5c40\u90e8\u7f51\u7edc\uff08cross-stage partial network, CSPNet\uff09[33] \u6765\u51cf\u5c11\u6a21\u578b\u5728\u63a8\u7406\u4e2d\u6240\u9700\u7684\u8ba1\u7b97\u91cf\u3002\u5934\u6a21\u5757\u7528\u4e8e\u6267\u884c\u5206\u7c7b\u548c\u56de\u5f52\u4efb\u52a1\uff0c\u91c7\u7528\u7279\u5f81\u91d1\u5b57\u5854\u7f51\u7edc\uff08feature pyramid network, FPN\uff09\u548c\u8def\u5f84\u805a\u5408\u7f51\u7edc\uff08path aggregation network, PAN\uff09[34]\u3002<\/p>\n<p>\u4e3a\u4e86\u63d0\u9ad8\u5bf9\u6781\u5c0f\u76ee\u6807\u533a\u57df\u7684\u68c0\u6d4b\u6548\u679c\uff0c\u8f93\u5165\u5c42\u91c7\u7528\u4e86mosaic\u6570\u636e\u589e\u5f3a\uff08mosaic data augmentation\uff09[35] \u65b9\u6cd5\uff0c\u5c06\u56db\u4e2a\u968f\u673a\u7f29\u653e\u3001\u526a\u5207\u548c\u968f\u673a\u6392\u5217\u7684\u56fe\u50cf\u62fc\u63a5\u5728\u4e00\u8d77\u3002\u6a21\u5757\u9aa8\u67b6\u5305\u62ec CSPNet \u548c\u7a7a\u95f4\u91d1\u5b57\u5854\u6c60\u5316\uff08spatial pyramid pooling, SPP\uff09[36] \u64cd\u4f5c\u3002\u8f93\u5165\u56fe\u50cf\u901a\u8fc7\u4e09\u4e2a CSP \u64cd\u4f5c\u548c\u4e00\u4e2a SPP \u64cd\u4f5c\uff0c\u751f\u6210\u4e86\u4e00\u4e2a\u56db\u500d\u4e8e\u539f\u59cb\u5927\u5c0f\u7684\u7279\u5f81\u56fe\u3002\u5934\u6a21\u5757\u6709\u4e09\u4e2a\u5206\u652f\u7f51\u7edc\uff0c\u5206\u522b\u63a5\u6536\u6765\u81ea\u4e0d\u540c\u5c42\u7684\u878d\u5408\u7279\u5f81\u3001\u8f93\u51fa\u5404\u5c42\u7684\u8fb9\u754c\u6846\u56de\u5f52\u503c\u548c\u76ee\u6807\u7c7b\u522b\uff0c\u6700\u540e\u7531\u5934\u6a21\u5757\u5408\u5e76\u5206\u652f\u7f51\u7edc\u7684\u9884\u6d4b\u7ed3\u679c\u3002<\/p>\n<p>\u56fe 1-5 \u76ee\u6807\u68c0\u6d4b\u7f51\u7edc\u7ed3\u6784<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-12.00.26\u202fPM.png\" alt=\"img\" \/><\/p>\n<p><strong>1.3. \u5b9e\u9a8c<\/strong><\/p>\n<p><strong>1.3.1. \u4e34\u5e8a\u4fe1\u606f\u548c\u5f71\u50cf\u6570\u636e\u96c6<\/strong><\/p>\n<p>\u672c\u9879\u76ee\u9009\u53d6\u7b26\u5408\u4ee5\u4e0b\u6761\u4ef6\u7684\u60a3\u8005\u5f00\u5c55\u7814\u7a76\uff1a1\uff09\u4e8e 2010 \u5e74 1 \u6708\u81f3 2018 \u5e74 12\u6708\u671f\u95f4\u5728\u56fd\u5bb6\u764c\u75c7\u4e2d\u5fc3\u4e2d\u56fd\u533b\u5b66\u79d1\u5b66\u9662\u80bf\u7624\u533b\u9662\u88ab\u8bca\u65ad\u4e3a\u539f\u53d1\u6027\u5bab\u9888\u764c\u7684\u60a3\u8005 2\uff09\u6709FDG-PET\/CT \u56fe\u50cf\uff1b3\uff09\u6709\u7535\u5b50\u75c5\u5386\u8bb0\u5f55\u3002\u603b\u5171\u5165\u7ec4\u4e86 220 \u540d\u60a3\u8005\uff0c\u5171\u8ba1 72,602 \u5f20\u5207\u7247\u56fe\u50cf\uff0c\u5e73\u5747\u6bcf\u4f4d\u60a3\u8005\u6709 330 \u5f20\u5207\u7247\u56fe\u50cf\u5165\u7ec4\u5b9e\u9a8c\u3002\u5176\u4e2d\uff0cCT \u5207\u7247\u56fe\u50cf\u7684\u9ad8\u5ea6\u548c\u5bbd\u5ea6\u5747\u4e3a512 \u50cf\u7d20\uff0c\u800cPET \u5207\u7247\u56fe\u50cf\u7684\u9ad8\u5ea6\u548c\u5bbd\u5ea6\u5747\u4e3a128 \u50cf\u7d20\uff0c\u6bcf\u4e2a\u6a21\u6001\u7684\u6570\u636e\u96c6\u90fd\u5305\u542b\u4e866,378\u5f20\u5207\u7247\u56fe\u50cf\uff0c\u5373\u5e73\u5747\u6bcf\u4f4d\u60a3\u8005\u670929\u5f20\u5207\u7247\u56fe\u50cf\uff0c\u7528\u4e8e\u8bad\u7ec3\u548c\u6d4b\u8bd5\u3002\u5728\u5165\u7ec4\u8fdb\u884c\u5206\u6790\u4e4b\u524d\uff0c\u6240\u6709\u60a3\u8005\u6570\u636e\u90fd\u5df2\u53bb\u6807\u8bc6\u5316\u3002\u672c\u7814\u7a76\u5df2\u83b7\u5f97\u5317\u4eac\u534f\u548c\u533b\u5b66\u9662\u56fd\u5bb6\u764c\u75c7\u4e2d\u5fc3\u4f26\u7406\u59d4\u5458\u4f1a\u7684\u6279\u51c6\u3002<\/p>\n<p>\u8be5\u6570\u636e\u96c6\u5305\u542b220 \u4e2a\u60a3\u8005\u7684\u5168\u8eab CT \u548c\u5168\u8eab PET \u56fe\u50cf\u6570\u636e\uff0c\u56e0\u5165\u7ec4\u7684\u6bcf\u4f4d\u60a3\u8005\u5747\u786e\u8bca\u4e3a\u5bab\u9888\u764c\uff0c\u6570\u636e\u96c6\u4e2d\u5404\u4f8b\u6570\u636e\u5747\u5305\u542b\u75c5\u53d8\u533a\u57df\uff0c\u5982\u8868 1-1 \u6240\u793a\u3002\u9274\u4e8e\u6240\u6709\u60a3\u8005\u7684CT \u548cPET \u5747\u5728\u540c\u4e00\u65f6\u95f4\u4e14\u4f7f\u7528\u76f8\u540c\u8bbe\u5907\u91c7\u96c6\uff0c\u56e0\u6b64 CT \u548cPET \u5c55\u793a\u7684\u89e3\u5256\u4fe1\u606f\u4e0e\u4ee3\u8c22\u4fe1\u606f\u6765\u81ea\u540c\u4e00\u65f6\u523b\u60a3\u8005\u8eab\u4f53\u7684\u540c\u4e00\u533a\u57df\uff0c\u5176\u7279\u5f81\u5177\u6709\u4e00\u5bf9\u4e00\u5bf9\u5e94\u4e14\u53ef\u5339\u914d\u7684\u7279\u6027\u3002\u6839\u636e\u80bf\u7624\u5927\u5c0f\u3001\u6d78\u6da6\u6df1\u5ea6\u3001\u76c6\u8154\u4e34\u8fd1\u7ec4\u7ec7\u4fb5\u72af\u7a0b\u5ea6\u3001\u8179\u76c6\u8154\u6dcb\u5df4\u7ed3\u8f6c\u79fb\u7684\u60c5\u51b5\u53ef\u5c06\u5bab\u9888\u764c\u7684\u8fdb\u5c55\u7a0b\u5ea6\u8fdb\u884c\u5206\u671f\uff0c\u4e3b\u8981\u5305\u62ec\u56db\u671f\uff0c\u6bcf\u671f\u4e2d\u53c8\u8fdb\u4e00\u6b65\u7ec6\u5206\u4e3a\u66f4\u5177\u4f53\u7684\u671f\u522b\u3002\u56fd\u9645\u5987\u4ea7\u79d1\u8054\u76df\uff08International Federation of Gynecology and Obstetrics,<br \/>\nFIGO\uff09\u4e8e 2018 \u5e74 10 \u6708\u66f4\u65b0\u4e86\u5bab\u9888\u764c\u5206\u671f\u7cfb\u7edf\u7684\u6700\u65b0\u7248\u672c[37]\u3002\u672c\u9879\u76ee\u6570\u636e\u96c6\u56ca\u62ec\u4e86 FIGO \u5206\u671f\u5168\u90e8\u56db\u4e2a\u671f\u522b\u7684\u5bab\u9888\u764c\u5f71\u50cf\u3002\u4e3a\u4e86\u4fdd\u6301\u8bad\u7ec3\u548c\u6d4b\u8bd5\u7684\u516c\u5e73\u6027\uff0c\u7eb3\u5165\u8bad\u7ec3\u96c6\u548c\u6d4b\u8bd5\u96c6\u7684\u4e0d\u540c\u671f\u522b\u5f71\u50cf\u7684\u5206\u5e03\uff0c\u5373\u4e0d\u540c FIGO \u5206\u671f\u7684\u5212\u5206\u6bd4\u4f8b\uff0c\u9700\u4fdd\u6301\u4e00\u81f4\uff0c\u5426\u5219\u53ef\u80fd\u4f1a\u5bfc\u81f4\u67d0\u4e9b FIGO \u671f\u522b\u7684\u6570\u636e\u96c6\u65e0\u6cd5\u8fdb\u884c\u8bad\u7ec3\u6216\u6d4b\u8bd5\u3002\u56e0\u6b64\uff0c\u5728\u4fdd\u8bc1\u5904\u4e8e\u4e0d\u540c\u671f\u522b\u7684\u60a3\u8005\u6570\u636e\u7684\u5212\u5206\u6bd4\u4f8b\u7684\u57fa\u7840\u4e0a\uff0c\u91c7\u7528\u4e94\u6298\u4ea4\u53c9\u9a8c\u8bc1\u65b9\u6cd5\u5c06220 \u540d\u60a3\u8005\u7684\u6570\u636e\u8fdb\u884c\u4e94\u7b49\u5206\uff0c\u6bcf\u4e2a\u90e8\u5206\u5927\u7ea6\u5305\u62ec\u4e86 45 \u4f8b\u60a3\u8005\u7684\u6570\u636e\uff0c\u5728\u6bcf\u8f6e\u9a8c\u8bc1\u4e2d\u968f\u673a\u9009\u62e9\u4e00\u4e2a\u90e8\u5206\u4f5c\u4e3a\u6d4b\u8bd5\u96c6\u3002\u6240\u6709\u6a21\u578b\u90fd\u9700\u8981\u8fdb\u884c5 \u6b21\u8bad\u7ec3\u548c\u8bc4\u4f30\uff0c\u4ee5\u83b7\u53d6\u5728\u6d4b\u8bd5\u96c6\u4e0a\u8868\u73b0\u51fa\u7684\u6027\u80fd\u7684\u5e73\u5747\u503c\u548c\u6807\u51c6\u5dee\u3002<\/p>\n<p>\u8868 1-1 \u6570\u636e\u96c6\u4e2d\u7684\u75c5\u4f8b\u6570\u53ca\u4e34\u5e8a\u5206\u671f<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-12.02.25\u202fPM.png\" alt=\"img\" \/><\/p>\n<p><strong>1.3.2. \u6a21\u578b\u8bad\u7ec3\u8fc7\u7a0b<\/strong><\/p>\n<p>\u5728\u6309\u4e0a\u8ff0\u6b65\u9aa4\u51c6\u5907\u597d\u6570\u636e\u96c6\u540e\uff0c\u9996\u5148\u5c06\u56fe\u50cf\u4ece 512\u00d7512 \u50cf\u7d20\u8c03\u6574\u4e3a 1024\u00d71024\u50cf\u7d20\uff0c\u7136\u540e\u4f7f\u7528\u591a\u79cd\u6570\u636e\u589e\u5f3a\u65b9\u6cd5\uff0c\u5305\u62ec mosaic \u589e\u5f3a[38]\u3001HSV\uff08Hue, Saturation, Value\uff09\u989c\u8272\u7a7a\u95f4\u589e\u5f3a[39]\u3001\u968f\u673a\u56fe\u50cf\u5e73\u79fb\u3001\u968f\u673a\u56fe\u50cf\u7f29\u653e\u548c\u968f\u673a\u56fe\u50cf\u7ffb\u8f6c\uff0c\u589e\u52a0\u8f93\u5165\u6570\u636e\u96c6\u5bf9\u566a\u58f0\u7684\u9c81\u68d2\u6027\u3002\u5728\u6bcf\u6b21\u5377\u79ef\u540e\u548c\u6fc0\u6d3b\u51fd\u6570\u524d\u8fdb\u884c\u6279\u5f52\u4e00\u5316\uff08Batch Normalization, BN\uff09[40] \u3002\u6240\u6709\u9690\u85cf\u5c42\u90fd\u91c7\u7528 Sigmoid \u52a0\u6743\u7ebf\u6027\u5355\u5143 (Sigmoid-Weighted Linear Units, SiLU)[41 \u4f5c\u4e3a\u6fc0\u6d3b\u51fd\u6570\u3002\u8bad\u7ec3\u6a21\u578b\u6240\u7528\u7684\u5b66\u4e60\u7387\u8bbe\u7f6e\u4e3a 1e-5\uff0c\u5e76\u5728\u8d77\u59cb\u8bad\u7ec3\u65f6\u9009\u62e9\u8f83\u5c0f\u7684\u5b66\u4e60\u7387\uff0c\u7136\u540e\u5728 5 \u4e2a\u8f6e\u6b21\uff08epoch\uff09\u540e\u4f7f\u7528\u9884\u8bbe\u7684\u5b66\u4e60<br \/>\n\u7387\u3002\u6bcf\u4e2a\u6a21\u578b\u4f7f\u7528PyTorch \u6846\u67b6\u57284 \u4e2aNvidia Tesla V100-SXM2 32G GPU \u4e0a\u8fdb\u884c50\u4e2a\u8f6e\u6b21\u7684\u8bad\u7ec3\u3002\u4f7f\u7528 0.98 \u7684\u52a8\u91cf\u548c 0.01 \u7684\u6743\u503c\u8870\u51cf\u901a\u8fc7\u968f\u673a\u68af\u5ea6\u4e0b\u964d\u6cd5\uff08Stochastic Gradient Descend, SGD\uff09\u6765\u4f18\u5316\u5404\u7f51\u7edc\u5c42\u7684\u6743\u91cd\u76ee\u6807\u51fd\u6570\u3002\u5728\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\uff0c\u7f51\u7edc\u5728\u9a8c\u8bc1\u96c6\u4e0a\u8fbe\u5230\u6700\u5c0f\u7684\u635f\u5931\u65f6\uff0c\u9009\u62e9\u6700\u4f73\u53c2\u6570\u3002\u6240\u6709\u5b9e\u9a8c\u4e2d\u7684\u6027\u80fd\u6d4b\u91cf\u90fd\u662f\u5728\u91c7\u7528\u6700\u4f18\u53c2\u6570\u8bbe\u7f6e\u7684\u6a21\u578b\u4e2d\u5bf9\u6d4b\u8bd5\u96c6\u8fdb\u884c\u6d4b\u8bd5\u5f97\u5230\u7684\uff0c\u8be6\u89c1\u8868 1-2\u3002<\/p>\n<p>\u4e3a\u4e86\u8fdb\u4e00\u6b65\u8bc1\u660e\u672c\u9879\u76ee\u6240\u63d0\u51fa\u7684\u6a21\u578b\u7684\u666e\u9002\u6027\uff0c\u9009\u62e9\u4e86\u516d\u4e2a\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u7684\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u4f5c\u4e3a\u57fa\u51c6\uff0c\u5e76\u6d4b\u8bd5\u4e86\u6240\u6709\u6a21\u578b\u5728\u8f93\u5165\u4e0d\u540c\u7684\u56fe\u50cf\u878d\u5408\u7ed3\u679c\u65f6\u7684\u6027\u80fd\u3002\u6bcf\u4e2a\u6a21\u578b\u7684\u8f93\u5165\u5b8c\u5168\u76f8\u540c\uff0c\u800c\u552f\u4e00\u7684\u533a\u522b\u662f\u795e\u7ecf\u7f51\u7edc\u4e2d\u7684\u8d85\u53c2\u6570\u6765\u81ea\u6bcf\u4e2a\u6a21\u578b\u7684\u5b98\u65b9\u8bbe\u7f6e\uff0c\u800c\u8fd9\u4e9b\u8d85\u53c2\u6570\u56e0\u6a21\u578b\u800c\u5f02\u3002<\/p>\n<p>\u8868 1-2 \u7f51\u7edc\u8bad\u7ec3\u7684\u8d85\u53c2\u6570<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-12.03.34\u202fPM.png\" alt=\"img\" \/><\/p>\n<p><strong>1.3.3. \u8bc4\u4ef7\u6307\u6807<\/strong><\/p>\n<p>\u672c\u9879\u76ee\u4f7f\u7528\u201c\u51c6\u786e\u5ea6\u7cbe\u503c50\u201d\uff08Accuracy Precision 50, AP50\uff09\u6765\u8bc4\u4f30\u76ee\u6807\u68c0\u6d4b\u7684\u6027\u80fd\u3002AP50 \u662f\u5f53\u4ea4\u5e76\u6bd4\uff08Intersection over Union, IOU\uff09\u9608\u503c\u4e3a0.5 \u65f6\u7684\u5e73\u5747\u7cbe\u5ea6\uff0c\u5982\u516c\u5f0f3\u6240\u5b9a\u4e49\uff0c\u5176\u4e2dP\u548cR\u5206\u522b\u662f\u7cbe\u5ea6\uff08Precision\uff09\u548c\u53ec\u56de\u7387\uff08Recall\uff09\u7684\u7f29\u5199\u3002\u6a21\u578b\u7684\u9884\u6d4b\u7ed3\u679c\u4f1a\u6709\u4e0d\u540c\u7684\u53ec\u56de\u7387\u548c\u7cbe\u5ea6\u503c\uff0c\u8fd9\u53d6\u51b3\u4e8e\u7f6e\u4fe1\u5ea6\u9608\u503c\u3002\u5c06\u53ec\u56de\u7387\u4f5c\u4e3a\u6a2a\u8f74\uff0c\u7cbe\u5ea6\u4f5c\u4e3a\u7eb5\u8f74\uff0c\u53ef\u4ee5\u7ed8\u5236 PR \u66f2\u7ebf\uff0c\u800c AP \u662f\u8be5\u66f2\u7ebf\u4e0b\u7684\u9762\u79ef\u3002IOU \u662f\u5c06\u771f\u5b9e\u6807\u6ce8\u533a\u57df\u548c\u6a21\u578b\u9884\u6d4b\u533a\u57df\u7684\u91cd\u53e0\u90e8\u5206\u9664\u4ee5\u4e24\u533a\u57df\u7684\u96c6\u5408\u90e8\u5206\uff08\u5373\u771f\u5b9e\u533a\u57df\u548c\u9884\u6d4b\u533a\u57df\u7684\u5e76\u96c6\uff09\u5f97\u5230\u7684\u7ed3\u679c\uff0c\u5982\u516c\u5f0f4 \u6240\u793a\u3002\u7cbe\u5ea6\u548c\u53ec\u56de\u7387\u7684\u8ba1\u7b97\u65b9\u5f0f\u5206\u522b\u5728\u516c\u5f0f1 \u548c2 \u4e2d\u5217\u51fa\uff0c\u5176\u4e2d\u771f\u6b63\u4f8b\u8868\u793a\u9884\u6d4b\u4e3a\u6b63\u4f8b\u7684\u6b63\u6837\u672c\uff0c\u5047\u6b63\u4f8b\u548c\u5047\u8d1f\u4f8b\u4ee3\u8868\u7684\u6982\u5ff5\u4ee5\u6b64\u7c7b\u63a8\u3002\u7cbe\u5ea6\u8868\u660e\u5728\u6a21\u578b\u9884\u6d4b\u7ed3\u679c\u91cc\uff0c\u88ab\u5224\u65ad\u4e3a\u6b63\u4f8b\u7684\u6837\u672c\u4e2d\u6709\u591a\u5c11\u5b9e\u9645\u662f\u6b63\u4f8b\uff0c\u800c\u53ec\u56de\u7387\u8868\u793a\u5b9e\u9645\u4e3a\u6b63\u4f8b\u7684\u6837\u672c\u4e2d\u591a\u5c11\u88ab\u9884\u6d4b\u4e3a\u6b63\u4f8b\u3002\u8868 1-3 \u8bb0\u5f55\u4e86\u56fe\u50cf\u6570\u636e\u96c6\u4ea4\u53c9\u9a8c\u8bc1\u540e\u5404\u4e2a\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u7684 AP50 \u7684\u5e73\u5747\u503c\u548c\u65b9\u5dee\u3002<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-12.04.55\u202fPM.png\" alt=\"img\" \/><\/p>\n<p><strong>1.3.4. \u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u7684\u7ed3\u679c\u4e0e\u5206\u6790<\/strong><\/p>\n<p>\u672c\u9879\u76ee\u91c7\u7528\u4e0d\u540c\u7684\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\uff0c\u5305\u62ec\u5355\u9636\u6bb5\u76ee\u6807\u68c0\u6d4b\u6a21\u578b(YOLOv5 [32]\u3001RetinaNet[42]\u3001ATSS [43 )\u548c\u4e8c\u9636\u6bb5\u76ee\u6807\u68c0\u6d4b\u6a21\u578b(Faster-RCNN [44<br \/>\n\u3001CascadeRCNN [45]\u3001LibraRCNN[46])\uff0c\u5728\u4e94\u6298\u4ea4\u53c9\u9a8c\u8bc1\u4e0b\u6bd4\u8f83\u4e86\u4f7f\u7528 CT \u56fe\u50cf\u3001PET \u56fe\u50cf\u3001PET-CT \u7b80\u5355\u878d\u5408\u56fe\u50cf\u3001PET-CT \u901a\u9053\u62fc\u63a5\u878d\u5408\u56fe\u50cf\uff08concat fusion\uff09\u548c\u672c\u9879\u76ee\u6240\u63d0\u51fa\u7684 PET-CT\u81ea\u9002\u5e94\u533a\u57df\u7279\u5f81\u878d\u5408\u56fe\u50cf\u4f5c\u4e3a\u8f93\u5165\u6570\u636e\u96c6\u65f6\uff0c\u6bcf\u4e2a\u6a21\u578b\u7684\u76ee\u6807\u68c0\u6d4b\u6027\u80fd\u3002\u5176\u4e2d\uff0cCT\u548cPET\u662f\u5355\u6a21\u6001\u56fe\u50cf\uff0c\u800cPET-CT\u7b80\u5355\u878d\u5408\u56fe\u50cf\u3001PET-CT\u901a\u9053\u62fc\u63a5\u56fe\u50cf\u548cPET-CT \u81ea\u9002\u5e94\u533a\u57df\u7279\u5f81\u878d\u5408\u56fe\u50cf\u662f\u8de8\u6a21\u6001\u878d\u5408\u56fe\u50cf\u3002\u7b80\u5355\u878d\u5408\u662f\u6307\u5c06 PET \u56fe\u50cf\u7b80\u5355\u5730\u7f29\u653e\u5230\u4e0e CT \u56fe\u50cf\u76f8\u540c\u7684\u5927\u5c0f\u540e\u8fdb\u884c\u50cf\u7d20\u503c\u7684\u53e0\u52a0\uff0c\u800c\u901a\u9053\u62fc\u63a5\u878d\u5408\u662f\u76f4\u63a5\u5c06\u4e24\u79cd\u6a21\u6001\u56fe\u50cf\u5728\u901a\u9053\u4e0a\u4e32\u8054\u5728\u4e00\u8d77\u4f5c\u4e3a\u7f51\u7edc\u7684\u8f93\u5165\u3002<\/p>\n<p>\u5982\u8868 1-3 \u6240\u793a\uff0c\u52a0\u7c97\u7684\u6570\u5b57\u4ee3\u8868\u6bcf\u884c\u4e2d\u6700\u597d\u7684\u5b9e\u9a8c\u7ed3\u679c\u3002\u4e0e\u4f7f\u7528\u5355\u4e00\u6a21\u6001\u6570\u636e\u8fdb\u884c\u80bf\u7624\u68c0\u6d4b\u6a21\u578b\u5206\u6790\uff08\u5982\u53ea\u4f7f\u7528CT \u6216PET \u56fe\u50cf\uff09\u76f8\u6bd4\uff0c\u672c\u9879\u76ee\u6240\u63d0\u51fa\u7684\u81ea\u9002\u5e94\u8de8\u6a21\u6001\u56fe\u50cf\u878d\u5408\u65b9\u6cd5\u5728\u76ee\u6807\u68c0\u6d4b\u4efb\u52a1\u4e2d\u5c55\u73b0\u51fa\u4e86\u66f4\u9ad8\u7684\u68c0\u6d4b\u7cbe\u5ea6\u3002\u7531\u4e8e\u81ea\u9002\u5e94\u878d\u5408\u65b9\u6cd5\u80fd\u591f\u5728\u8de8\u6a21\u6001\u878d\u5408\u4e4b\u524d\u5c06\u4e24\u79cd\u6a21\u6001\u56fe\u50cf\u7684\u4fe1\u606f\u8fdb\u884c\u9884\u5bf9\u9f50\uff0c\u5bf9 CT \u56fe\u50cf\u548cPET\u56fe\u50cf\u7684\u7ed3\u6784\u7279\u5f81\u8fdb\u884c\u4e00\u4e00\u914d\u51c6\uff0c\u56e0\u6b64\uff0c\u4e0e\u7b80\u5355\u878d\u5408\u65b9\u6cd5\u548c\u901a\u9053\u62fc\u63a5\u878d\u5408\u65b9\u6cd5\u76f8\u6bd4\uff0c\u81ea\u9002\u5e94\u878d\u5408\u65b9\u6cd5\u7684\u6027\u80fd\u6700\u4f73\u3002\u4e0a\u8ff0\u9488\u5bf9\u4e0d\u540c\u6a21\u6001\u56fe\u50cf\u53ca\u4f7f\u7528\u4e0d\u540c\u8de8\u6a21\u6001\u878d\u5408\u65b9\u6cd5\u4f5c\u4e3a\u8f93\u5165\u5f97\u5230\u7684\u6d4b\u8bd5\u6027\u80fd\u7ed3\u679c\u5728\u4f7f\u7528\u4e0d\u540c\u7684\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u7684\u60c5\u51b5\u4e0b\u4fdd\u6301\u4e00\u81f4\uff0c\u8fd9\u8868\u660e\u672c\u9879\u76ee\u6240\u63d0\u51fa\u7684\u8de8\u6a21\u6001\u81ea\u9002\u5e94\u878d\u5408\u65b9\u6cd5\u6709\u826f\u597d\u7684\u901a\u7528\u6027\uff0c\u53ef\u4ee5\u6cdb\u5316\u5e94\u7528\u5230\u5404\u79cd\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u7684\u9884\u5904\u7406\u4e2d\u3002<\/p>\n<p>\u8868 1-3 \u4e94\u6298\u4ea4\u53c9\u9a8c\u8bc1\u76ee\u6807\u68c0\u6d4b\u5b9e\u9a8c\u7684\u7ed3\u679c\uff08\u201c*\u201d \u8868\u793a\u4ea4\u53c9\u9a8c\u8bc1\u4e2d\u7684\u67d0\u4e00\u6298\u5728\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u51fa\u73b0\u68af\u5ea6\u7206\u70b8\uff0c\u6570\u503c\u4e3a\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u7684 AP50 \u7684\u5747\u503c\u548c\u65b9\u5dee\uff09<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-12.07.42\u202fPM.png\" alt=\"img\" \/><\/p>\n<p>\u56fe 1-6 \u5c06\u4e0d\u540c\u6a21\u6001\u56fe\u50cf\u4e0b\u76ee\u6807\u5bab\u9888\u764c\u75c5\u53d8\u533a\u57df\u7684\u68c0\u6d4b\u7ed3\u679c\u548c\u5b9e\u9645\u6807\u6ce8\u7684\u764c\u7076\u533a\u57df\u8fdb\u884c\u4e86\u53ef\u89c6\u5316\u3002\u5176\u4e2d\u7eff\u8272\u6846\u662f\u7531\u533b\u5e08\u6807\u6ce8\u7684\u771f\u5b9e\u75c5\u53d8\u533a\u57df\uff0c\u9ec4\u8272\u6846\u662f\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u7684\u9884\u6d4b\u7ed3\u679c\u3002\u5206\u6790\u56fe\u50cf\u6a21\u6001\u4fe1\u606f\u53ef\u77e5\uff0cCT \u56fe\u50cf\u65e2\u5305\u542b\u4e86\u4eba\u4f53\u6b63\u5e38\u7ed3\u6784\u7684\u4fe1\u606f\uff0c\u4e5f\u5305\u542b\u75c5\u7076\u7684\u89e3\u5256\u4fe1\u606f\uff0c\u524d\u8005\u53ef\u80fd\u4f1a\u5e72\u6270\u5bab\u9888\u764c\u75c5\u53d8\u533a\u57df\u7279\u5f81\u7684\u8bc6\u522b\u548c\u68c0\u6d4b\u3002\u56e0\u6b64\uff0c\u5728\u5355\u4e00 CT \u6a21\u6001\u4e0b\u4f1a\u6709\u4e00\u4e9b\u6f0f\u68c0\u3002\u4e0e CT \u6a21\u6001\u7684\u9884\u6d4b\u6846\u76f8\u6bd4\uff0cPET \u6a21\u6001\u4e0b\u7684\u9884\u6d4b\u6846\u4e0e\u6807\u6ce8\u6846\u7684 IOU \u66f4\u9ad8\uff0c\u6216\u8bb8\u662f\u7531\u4e8e PET \u5f71\u50cf\u6709\u66f4\u591a\u80fd\u8868\u73b0\u5bab\u9888\u764c\u533a\u57df\u7279\u5f81\u7684\u4fe1\u606f\u3002\u5728 PET-CT \u533a\u57df\u7279\u5f81\u8de8\u6a21\u6001\u878d\u5408\u56fe\u50cf\u4e2d\u68c0\u6d4b\u6548\u679c\u6700\u4f73\uff0c\u56e0\u4e3a PET-CT \u878d\u5408\u56fe\u50cf\u878d\u5408\u4e86\u4e24\u79cd\u6a21\u6001\u7684\u4e0d\u540c\u7279\u5f81\uff0c\u4ece\u800c\u5927\u5927\u63d0\u9ad8\u68c0\u6d4b\u7684\u51c6\u786e\u6027\u3002<\/p>\n<p>\u56fe 1-6 \u8de8\u6a21\u6001\u878d\u5408\u56fe\u50cf\u7684\u76ee\u6807\u68c0\u6d4b\u7ed3\u679c<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-12.08.38\u202fPM.png\" alt=\"img\" \/><\/p>\n<p><strong>1.4. \u8ba8\u8bba<\/strong><\/p>\n<p>\u672c\u9879\u76ee\u65e8\u5728\u8bc4\u4f30\u6df1\u5ea6\u5b66\u4e60\u7b97\u6cd5\u662f\u5426\u53ef\u4ee5\u8de8\u6a21\u6001\u878d\u5408 FDG-PET \u548c CT \u56fe\u50cf\uff0c\u5e76\u5728\u878d\u5408\u56fe\u50cf\u4e2d\u5b9e\u73b0\u5bab\u9888\u764c\u75c5\u7076\u533a\u57df\u7684\u81ea\u52a8\u68c0\u6d4b\u3002\u6211\u4eec\u63d0\u51fa\u4e86\u4e00\u4e2a\u57fa\u4e8e\u8de8\u6a21\u6001\u878d\u5408\u56fe\u50cf\u7684\u68c0\u6d4b\u6846\u67b6\uff0c\u5305\u62ec\u533a\u57df\u7279\u5f81\u5339\u914d\u3001\u56fe\u50cf\u878d\u5408\u548c\u76ee\u6807\u68c0\u6d4b\u7b49\u6b65\u9aa4\u3002\u878d\u5408 CT \u548cPET \u56fe\u50cf\u53ef\u4ee5\u6700\u5927\u7a0b\u5ea6\u5730\u63d0\u53d6\u5404\u4e2a\u6a21\u6001\u4e2d\u5305\u542b\u7684\u4fe1\u606f\uff0c\u56e0\u6b64 PET-CT \u8de8\u6a21\u6001\u878d\u5408\u56fe\u50cf\u542b\u6709\u4e30\u5bcc\u7684\u89e3\u5256\u548c\u529f\u80fd\u4fe1\u606f\u3002\u76ee\u6807\u68c0\u6d4b\u5b9e\u9a8c\u8bc1\u660e\uff0c\u672c\u9879\u76ee\u63d0\u51fa\u7684\u8de8\u6a21\u6001\u878d\u5408\u65b9\u6cd5\u5f97\u5230\u7684\u878d\u5408\u56fe\u50cf\u663e\u8457\u63d0\u9ad8\u4e86\u76ee\u6807\u68c0\u6d4b\u7684\u51c6\u786e\u6027\uff0c\u76f8\u6bd4\u5355\u6a21\u6001\u548c\u5176\u4ed6\u878d\u5408\u65b9\u6cd5\u5f97\u5230\u7684\u591a\u6a21\u6001\u56fe\u50cf\uff0c\u76ee\u6807\u68c0\u6d4b\u5e73\u5747\u7cbe\u786e\u5ea6\u5206\u522b\u63d0\u9ad8\u4e86 6.06%\u548c8.9%\u3002<\/p>\n<p>\u8868 1-3 \u5c55\u793a\u4e86\u57fa\u4e8e\u4e0d\u540c\u7684\u56fe\u50cf\u878d\u5408\u65b9\u6cd5\u5f62\u6210\u7684\u591a\u6a21\u6001\u56fe\u50cf\uff0c\u4e0d\u540c\u68c0\u6d4b\u6a21\u578b\u5728\u4e94\u6298\u4ea4\u53c9\u9a8c\u8bc1\u4e0b\u7684\u7ed3\u679c\u3002\u56e0\u5728\u89e3\u5256\u548c\u529f\u80fd\u5f71\u50cf\u4e2d\u5747\u6709\u5f02\u5e38\u8868\u73b0\u7684\u533a\u57df\u66f4\u53ef\u80fd\u662f\u764c\u53d8\uff0c\u6211\u4eec\u63a8\u6d4b\uff0c\u56fe\u50cf\u4fe1\u606f\u5bf9\u9f50\u6709\u5229\u4e8e\u5bf9\u5bab\u9888\u764c\u75c5\u7076\u7684\u76ee\u6807\u533a\u57df\u68c0\u6d4b\u3002\u56fe 1-6 \u5c55\u793a\u4e86\u5728\u4e0d\u540c\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u548c\u4e0d\u540c\u8f93\u5165\u56fe\u50cf\u6570\u636e\u6a21\u6001\u4e0b\u76ee\u6807\u68c0\u6d4b\u6548\u679c\u7684\u53ef\u89c6\u5316\u56fe\u50cf\u3002\u57fa\u4e8e\u672c\u9879\u76ee\u63d0\u51fa\u7684\u8de8\u6a21\u6001\u878d\u5408\u65b9\u6cd5\u751f\u6210\u7684\u56fe\u50cf\u8fdb\u884c\u7684\u76ee\u6807\u68c0\u6d4b\u7684\u68c0\u6d4b\u7ed3\u679c\u66f4\u4e3a\u51c6\u786e\uff0c\u5e76\u6d88\u9664\u4e86\u4e00\u4e9b\u5047\u9633\u6027\u7ed3\u679c\u3002\u6839\u636e\u533b\u751f\u7684\u65e5\u5e38\u8bca\u65ad\u4e60\u60ef\uff0c\u751f\u6210\u4e86\u4ee5\u7ea2\u8272\u548c\u9ec4\u8272\u4e3a\u4e3b\u8272\u7684\u878d\u5408\u56fe\u50cf\u3002<\/p>\n<p>\u5229\u7528 FDG-PET\/CT \u5bf9\u5bab\u9888\u764c\u8fdb\u884c\u53ca\u65f6\u3001\u51c6\u786e\u7684\u5206\u671f\u80fd\u591f\u5f71\u54cd\u60a3\u8005\u7684\u4e34\u5e8a\u6cbb\u7597\u51b3\u7b56\uff0c\u8fdb\u800c\u5ef6\u7f13\u75be\u75c5\u8fdb\u5c55\uff0c\u5e76\u51cf\u5c11\u80bf\u7624\u6cbb\u7597\u76f8\u5173\u7684\u6574\u4f53\u8d22\u52a1\u8d1f\u62c5[47] \u3002\u5bf9 FDG-PET\/CT \u56fe\u50cf\u7684\u89e3\u91ca\u5728\u5f88\u5927\u7a0b\u5ea6\u4e0a\u4f9d\u8d56\u4e34\u5e8a\u4e0a\u83b7\u5f97\u7684\u80cc\u666f\u4fe1\u606f\uff0c\u5e76\u9700\u8981\u7efc\u5408\u4e34\u5e8a\u5206\u6790\u6765\u786e\u5b9a\u662f\u5426\u53d1\u751f\u764c\u75c7\u7684\u6d78\u6da6\u548c\u8f6c\u79fb[48]\u3002\u5728\u67d0\u4e9b\u60c5\u51b5\u4e0b\uff0c\u6838\u533b\u5b66\u79d1\u9605\u7247\u533b\u5e08\u53ef\u4ee5\u8fc5\u901f\u8bc6\u522b\u5c40\u90e8\u6269\u5c55\u548c\u6dcb\u5df4\u6813\u585e\u3002\u800c\u591a\u6570\u60c5\u51b5\u4e0b\uff0c\u6838\u533b\u5b66\u79d1\u533b\u5e08\u5206\u6790\u4e00\u4f4d\u60a3\u8005\u7684FDG-PET\/CT \u5f71\u50cf\u5b66\u68c0\u67e5\u7ed3\u679c\u5e73\u5747\u9700\u8981\u4e09\u4e2a\u5c0f\u65f6\u3002\u6bd4\u8d77\u5360\u7528\u533b\u5e08\u6602\u8d35\u4e14\u7a00\u7f3a\u7684\u65f6\u95f4\uff0c\u5229\u7528\u8ba1\u7b97\u673a\u8fdb\u884c\u6b64\u9879\u5de5\u4f5c\u65e2\u80fd\u8282\u7ea6\u6210\u672c\uff0c\u9884\u8ba1\u8017\u65f6\u53c8\u77ed\uff0c\u4e14\u53ef\u4ee5\u5168\u5929\u5019\u8fd0\u884c\u3002\u672c\u9879\u76ee\u7684\u76ee\u6807\u662f\u901a\u8fc7\u4eba\u5de5\u667a\u80fd\u65b9\u6cd5\u5b9e\u73b0PET \u548cCT \u56fe\u50cf\u7684\u81ea\u52a8\u878d\u5408\uff0c\u5e76\u5229\u7528\u76ee\u6807\u68c0\u6d4b\u6280\u672f\u8bc6\u522b\u5bab\u9888\u764c\u7684\u6d78\u6da6\u548c\u8f6c\u79fb\uff0c\u4f5c\u4e3a\u8f85\u52a9\u5de5\u5177\u52a0\u901f FDG-PET\/CT \u7684\u9605\u7247\u8fc7\u7a0b\uff0c\u4ece\u800c\u4f7f\u4e34\u5e8a\u533b\u751f\u80fd\u591f\u5728\u6700\u77ed\u7684\u65f6\u95f4\u5185\u6309\u7167 FIGO \u6307\u5357\u5bf9\u5bab\u9888\u764c\u8fdb\u884c\u5206\u671f\u3002<\/p>\n<p>\u8fd9\u9879\u7814\u7a76\u4ecd\u5b58\u5728\u4e00\u4e9b\u5c40\u9650\u6027\u3002\u867d\u7136\u672c\u9879\u76ee\u5bf9\u57fa\u4e8e PET-CT \u81ea\u9002\u5e94\u878d\u5408\u56fe\u50cf\u7684\u76ee\u6807\u68c0\u6d4b\u65b9\u6cd5\u4e0e\u5176\u4ed6\u6700\u5148\u8fdb\u7684\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u7684\u76ee\u6807\u68c0\u6d4b\u65b9\u6cd5\u8fdb\u884c\u4e86\u6bd4\u8f83\uff0c\u4f46\u5c06\u8be5\u65b9\u6cd5\u62d3\u5c55\u5e94\u7528\u5230\u5176\u4ed6\u75c5\u79cd\u7684\u5f71\u50cf\u5b66\u5206\u6790\u7684\u53ef\u884c\u6027\u4ecd\u9700\u8bc4\u4f30\u3002\u6b64\u5916\uff0c\u6211\u4eec\u63d0\u51fa\u7684\u8de8\u6a21\u6001\u878d\u5408\u6846\u67b6\u5728\u56fe\u50cf\u878d\u5408\u65f6\u5e76\u672a\u8003\u8651\u6bcf\u79cd\u6a21\u6001\u56fe\u50cf\u7684\u6743\u91cd\u5206\u5e03\u3002\u672a\u6765\u53ef\u4ee5\u8bbe\u8ba1\u4e00\u79cd\u7279\u6b8a\u7684\u635f\u5931\u51fd\u6570\u6765\u8c03\u6574 ROI \u5185\u6bcf\u4e2a\u50cf\u7d20\u7684\u6743\u91cd\u5206\u5e03\uff0c\u4ee5\u63d0\u9ad8\u76ee\u6807\u68c0\u6d4b\u7ed3\u679c\u7684\u51c6\u786e\u6027\u3002<\/p>\n<p><strong>1.5. \u7ed3\u8bba<\/strong><\/p>\n<p>\u672c\u9879\u76ee\u63d0\u51fa\u4e86\u4e00\u79cd\u57fa\u4e8e\u8de8\u6a21\u6001\u56fe\u50cf\u878d\u5408\u7684\u591a\u6a21\u6001\u56fe\u50cf\u8fdb\u884c\u75c5\u53d8\u533a\u57df\u68c0\u6d4b\u7684\u6df1\u5ea6\u5b66\u4e60\u6846\u67b6\uff0c\u7528\u4e8e\u5bab\u9888\u764c\u7684\u68c0\u6d4b\u3002\u4e3a\u4e86\u5e94\u5bf9\u533b\u5b66\u5f71\u50cf\u4e2d\u5355\u4e00\u6a21\u6001\u56fe\u50cf\u5728\u80bf\u7624\u68c0\u6d4b\u65b9\u9762\u7684\u6027\u80fd\u4e0d\u8db3\uff0c\u63d0\u51fa\u4e86\u4e00\u79cd\u57fa\u4e8e\u533a\u57df\u7279\u5f81\u5339\u914d\u7684\u81ea\u9002\u5e94\u8de8\u6a21\u6001\u56fe\u50cf\u878d\u5408\u7b56\u7565\uff0c\u5c06\u878d\u5408\u540e\u7684\u591a\u6a21\u6001\u533b\u5b66\u56fe\u50cf\u8f93\u5165\u6df1\u5ea6\u5b66\u4e60\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u5b8c\u6210\u5bab\u9888\u764c\u75c5\u53d8\u533a\u57df\u68c0\u6d4b\u4efb\u52a1\uff0c\u5e76\u8ba8\u8bba\u4e86\u6df1\u5ea6\u5b66\u4e60\u6a21\u578b\u5728\u6bcf\u79cd\u6a21\u6001\u56fe\u50cf\u8f93\u5165\u95f4\u7684\u6027\u80fd\u5dee\u5f02\u3002\u5927\u91cf\u7684\u5b9e\u9a8c\u8bc1\u660e\uff0c\u4e0e\u4f7f\u7528\u5355\u4e00\u6a21\u6001\u7684\u5f71\u50cf\u53ca\u57fa\u4e8e\u7b80\u5355\u878d\u5408\u65b9\u6cd5\u6216\u901a\u9053\u62fc\u63a5\u878d\u5408\u65b9\u6cd5\u7684\u591a\u6a21\u6001\u5f71\u50cf\u76f8\u6bd4\uff0c\u81ea\u9002\u5e94\u878d\u5408\u540e\u7684\u591a\u6a21\u6001\u533b\u5b66\u56fe\u50cf\u66f4\u6709\u52a9\u4e8e\u5bab\u9888\u764c\u75c5\u53d8\u533a\u57df\u7684\u68c0\u6d4b\u3002<\/p>\n<p>\u672c\u9879\u76ee\u6240\u63d0\u51fa\u7684\u6280\u672f\u53ef\u5b9e\u73b0 PET \u548cCT \u56fe\u50cf\u7684\u81ea\u52a8\u878d\u5408\uff0c\u5e76\u5bf9\u5bab\u9888\u764c\u75c5\u53d8\u533a\u57df\u8fdb\u884c\u68c0\u6d4b\uff0c\u4ece\u800c\u8f85\u52a9\u533b\u751f\u7684\u8bca\u65ad\u8fc7\u7a0b\uff0c\u5177\u5907\u5b9e\u9645\u5e94\u7528\u4ef7\u503c\u3002\u540e\u7eed\u5c06\u57fa\u4e8e\u7b2c\u4e00\u90e8\u5206\u7684\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u57fa\u7840\uff0c\u5229\u7528\u7279\u5f81\u8f6c\u6362\u7684\u65b9\u6cd5\uff0c\u5c06\u56fe\u50cf\u6570\u636e\u8f6c\u6362\u4e3a\u7ed3\u6784\u6570\u636e\uff0c\u5c06\u8de8\u6a21\u6001\u878d\u5408\u65b9\u6cd5\u5e94\u7528\u4e8e\u5206\u7c7b\u95ee\u9898\u3002<\/p>\n<p><strong>2. \u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u6570\u636e\u878d\u5408\u7684\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u7684\u8bca\u65ad<\/strong><\/p>\n<p><strong>2.1. \u524d\u8a00<\/strong><\/p>\n<p>\u9aa8\u9abc\u662f\u7b2c\u4e09\u5e38\u89c1\u7684\u6076\u6027\u80bf\u7624\u8f6c\u79fb\u90e8\u4f4d\uff0c\u5176\u53d1\u751f\u7387\u4ec5\u6b21\u4e8e\u80ba\u8f6c\u79fb\u548c\u809d\u8f6c\u79fb\uff0c\u8fd1 70%\u7684\u9aa8\u8f6c\u79fb\u7624\u7684\u539f\u53d1\u90e8\u4f4d\u4e3a\u4e73\u817a\u548c\u524d\u5217\u817a[49] ,[50] \u3002\u9aa8\u8f6c\u79fb\u9020\u6210\u7684\u9aa8\u76f8\u5173\u4e8b\u4ef6\u975e\u5e38\u591a\u6837\uff0c\u4ece\u5b8c\u5168\u65e0\u75c7\u72b6\u5230\u4e25\u91cd\u75bc\u75db\u3001\u5173\u8282\u6d3b\u52a8\u5ea6\u964d\u4f4e\u3001\u75c5\u7406\u6027\u9aa8\u6298\u3001\u810a\u9ad3\u538b\u8feb\u3001\u9aa8\u9ad3\u8870\u7aed\u548c\u9ad8\u9499\u8840\u75c7\u3002\u9ad8\u9499\u8840\u75c7\u53c8\u53ef\u5bfc\u81f4\u4fbf\u79d8\u3001\u5c3f\u91cf\u8fc7\u591a\u3001\u53e3\u6e34\u548c\u75b2\u52b3\uff0c\u6216\u56e0\u8840\u9499\u6025\u5267\u5347\u9ad8\u5bfc\u81f4\u5fc3\u5f8b\u5931\u5e38\u548c\u6025\u6027\u80be\u529f\u80fd\u8870\u7aed[51 \u3002\u9aa8\u8f6c\u79fb\u662f\u4e73\u817a\u764c\u6700\u5e38\u89c1\u7684\u8f6c\u79fb\u65b9\u5f0f\uff0c\u4e5f\u662f\u60a3\u8005\u9884\u540e\u7684\u5206\u6c34\u5cad\uff0c\u5176\u8bca\u65ad\u540e\u7684\u4e2d\u4f4d\u751f\u5b58\u671f\u7ea6\u4e3a 40 \u4e2a\u6708[52] ,[53] \u3002\u56e0\u6b64\uff0c\u53ca\u65f6\u53d1\u73b0\u9aa8\u8f6c\u79fb\u75c5\u7076\u5bf9\u4e8e\u8bca\u65ad\u3001\u6cbb\u7597\u65b9\u6848\u7684\u9009\u62e9\u548c\u4e73\u817a\u764c\u60a3\u8005\u7684\u7ba1\u7406\u81f3\u5173\u91cd\u8981\u3002\u76ee\u524d\uff0c\u75c5\u7076\u7a7f\u523a\u6d3b\u68c0\u662f\u8bca\u65ad\u9aa8\u8f6c\u79fb\u7684\u91d1\u6807\u51c6\uff0c\u4f46\u9274\u4e8e\u7a7f\u523a\u6d3b\u68c0\u6709\u521b\u3001\u5b58\u5728\u8f83\u9ad8\u98ce\u9669\u3001\u4e14\u5047\u9634\u6027\u7387\u9ad8\uff0c\u4e34\u5e8a\u5e38\u7528\u5f71\u50cf\u5b66\u68c0\u67e5\u90e8\u5206\u66ff\u4ee3\u7a7f\u523a\u6d3b\u68c0\u5224\u65ad\u662f\u5426\u53d1\u751f\u9aa8\u8f6c\u79fb\u3002<\/p>\n<p>Batson\u7684\u7814\u7a76\u8868\u660e\uff0c\u4e73\u817a\u7684\u9759\u8109\u56de\u6d41\u4e0d\u4ec5\u6c47\u5165\u8154\u9759\u8109\uff0c\u8fd8\u6c47\u5165\u81ea\u9aa8\u76c6\u6cbf\u690e\u65c1\u8d70\u884c\u5230\u786c\u819c\u5916\u7684\u690e\u9759\u8109\u4e1b[54] \u3002\u901a\u8fc7\u690e\u9759\u8109\u4e1b\u5411\u9aa8\u9abc\u7684\u8840\u6db2\u56de\u6d41\u90e8\u5206\u89e3\u91ca\u4e86\u4e73\u817a\u764c\u6613\u5411\u4e2d\u8f74\u9aa8\u548c\u80a2\u5e26\u9aa8\u8f6c\u79fb\u7684\u539f\u56e0\u3002\u56e0\u6f5c\u5728\u9aa8\u8f6c\u79fb\u7076\u7684\u4f4d\u7f6e\u5206\u5e03\u8f83\u5e7f\uff0c\u5f71\u50cf\u5b66\u7b5b\u67e5\u9700\u8981\u8986\u76d6\u66f4\u5927\u7684\u533a\u57df\uff0c\u5e38\u8981\u6c42\u5168\u8eab\u663e\u50cf\u3002\u5e38\u7528\u7684\u9aa8\u8f6c\u79fb\u5f71\u50cf\u8bca\u65ad\u65b9\u6cd5\u5305\u62ec\u5168\u8eab\u9aa8\u663e\u50cf\uff08whole-body bone scintigraphy, WBS\uff09\u3001\u8ba1\u7b97\u673a\u65ad\u5c42\u626b\u63cf\uff08computed tomography, CT\uff09\u3001\u78c1\u5171\u632f\u6210\u50cf\uff08magnetic resonance imaging, MRI\uff09\u548c\u6b63\u7535\u5b50\u53d1\u5c04\u65ad\u5c42\u663e\u50cf\uff08positron emission tomography, PET\uff09[55]\u3002CT \u53ef\u4ee5\u6e05\u6670\u5730\u663e\u793a\u9aa8\u7834\u574f\uff0c\u786c\u5316\u6c89\u79ef\uff0c\u548c\u8f6c\u79fb\u7624\u5f15\u8d77\u8f6f\u7ec4\u7ec7\u80bf\u80c0\uff1bMRI \u5177\u6709\u4f18\u5f02\u7684\u9aa8\u548c\u8f6f\u7ec4\u7ec7\u5bf9\u6bd4\u5206\u8fa8\u7387\uff1b\u56e0 [18F] \u6c1f\u5316\u94a0\u4f1a\u7279\u5f02\u6027\u5730\u88ab\u9aa8\u7ec4\u7ec7\u5438\u6536\u3001\u4ee3\u8c22\uff0c PET \u53ef\u4ee5\u5b9a\u4f4d\u5168\u8eab\u5404\u5904\u9aa8\u4ee3\u8c22\u6d3b\u8dc3\u7684\u533a\u57df\u3002\u7136\u800c\uff0c\u5355\u4e00\u6a21\u6001\u5f71\u50cf\u5e38\u4e0d\u8db3\u4ee5\u68c0\u6d4b\u9aa8\u8f6c\u79fb\uff0c\u4e14\u7528\u4f20\u7edf\u65b9\u6cd5\u7efc\u5408\u5355\u4e00\u60a3\u8005\u7684 CT\u3001MRI\u3001PET \u6570\u636e\u7b5b\u67e5\u9aa8\u8f6c\u79fb\u75c5\u7076\u9700\u8981\u5bf9\u4e0a\u5343\u5e45\u5f71\u50cf\u8fdb\u884c\u89e3\u8bfb\uff0c\u8fd9\u4e00\u6781\u4e3a\u8017\u65f6\u7684\u8fc7\u7a0b\u53ef\u80fd\u5f71\u54cd\u4e34\u5e8a\u533b\u751f\u5bf9\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u7684\u8bca\u65ad\uff0c\u9020\u6210\u8bef\u8bca\u3001\u6f0f\u8bca\u3002\u800c\u9aa8\u8f6c\u79fb\u7684\u6f0f\u8bca\u4f1a\u8bef\u5bfc\u4e00\u7cfb\u5217\u4e34\u5e8a\u51b3\u7b56\uff0c\u5bfc\u81f4\u707e\u96be\u6027\u540e\u679c\u3002<\/p>\n<p>\u4f5c\u4e3a\u4e00\u79cd\u5ba2\u89c2\u8bc4\u4f30\u4f53\u7cfb\uff0c\u4eba\u5de5\u667a\u80fd\u8f85\u52a9\u9aa8\u8f6c\u79fb\u81ea\u52a8\u8bca\u65ad\u7cfb\u7edf\u901a\u8fc7\u51cf\u5c11\u89c2\u5bdf\u8005\u95f4\u548c\u89c2\u5bdf\u8005\u5185\u7684\u53d8\u5f02\u6027\uff0c\u63d0\u9ad8\u4e86\u8bca\u65ad\u7684\u4e00\u81f4\u6027\u548c\u53ef\u91cd\u590d\u6027\uff0c\u964d\u4f4e\u4e86\u5047\u9634\u6027\u7387\u3002\u5728\u51cf\u8f7b\u4e34\u5e8a\u533b\u5e08\u7684\u5de5\u4f5c\u8d1f\u62c5\u7684\u540c\u65f6\uff0c\u63d0\u9ad8\u8bca\u65ad\u7684\u51c6\u786e\u6027\u3002\u76ee\u524d\u5df2\u7ecf\u6709\u5f88\u591a\u5728\u5355\u4e00\u6a21\u6001\u56fe\u50cf\u4e2d\uff08CT\u3001MRI \u6216 PET\uff09\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u6280\u672f\u8fdb\u884c\u9aa8\u8f6c\u79fb\u75c5\u53d8\u68c0\u6d4b\u7684\u5de5\u4f5c\uff1a Noguchi \u7b49\u4eba\u5f00\u53d1\u4e86\u4e00\u79cd\u57fa\u4e8e\u6df1\u5ea6\u5b66\u4e60\u7684\u7b97\u6cd5\uff0c\u5b9e\u73b0\u4e86\u5728\u6240\u6709 CT \u626b\u63cf\u533a\u57df\u4e2d\u5bf9\u9aa8\u8f6c\u79fb\u75c5\u7076\u7684\u81ea\u52a8\u68c0\u6d4b [56]\uff1bFan \u7b49\u4eba\u7528 AdaBoost \u7b97\u6cd5\u548c Chan-Vese \u7b97\u6cd5\u5728 MRI \u56fe\u50cf\u4e0a\u5bf9\u80ba\u764c\u7684\u810a\u67f1\u8f6c\u79fb\u75c5\u7076\u8fdb\u884c\u4e86\u81ea\u52a8\u68c0\u6d4b\u548c\u5206\u5272\u80ba[57]\uff1bMoreau\u7b49\u4eba\u6bd4\u8f83\u4e86\u4e0d\u540c\u6df1\u5ea6\u5b66\u4e60\u7b97\u6cd5\u5728 PET\/CT \u56fe\u50cf\u4e0a\u5206\u5272\u6b63\u5e38\u9aa8\u7ec4\u7ec7\u548c\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u533a\u57df\u7684\u6027\u80fd[58] \u3002\u4f46\u5f88\u5c11\u6709\u4f7f\u7528\u8de8\u6a21\u6001\u6570\u636e\u878d\u5408\u7684\u6df1\u5ea6\u5b66\u4e60\u65b9\u6cd5\uff0c\u5224\u65ad\u662f\u5426\u5b58\u5728\u9aa8\u8f6c\u79fb\u7076\u7684\u76f8\u5173\u7814\u7a76\u3002<\/p>\n<p>\u65e8\u5728\u51cf\u8f7b\u4e34\u5e8a\u533b\u751f\u7684\u5de5\u4f5c\u8d1f\u62c5\uff0c\u672c\u7ae0\u63d0\u51fa\u4e86\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u6570\u636e\u878d\u5408\u65b9\u6cd5\uff0c\u7528\u4e8e\u5206\u6790 CT\u3001MRI \u548c PET \u56fe\u50cf\uff0c\u4ee5\u5224\u65ad\u5176\u4e2d\u662f\u5426\u5b58\u5728\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u75c5\u7076\u3002<\/p>\n<p>\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684 CT\u3001MRI \u548c PET \u8de8\u6a21\u6001\u56fe\u50cf\u6570\u636e\u878d\u5408\uff0c\u8fdb\u884c\u9aa8\u8f6c\u79fb\u75c5\u53d8\u5206\u7c7b\uff08\u5373\u5b58\u5728\u9aa8\u8f6c\u79fb\u75c5\u7076\u548c\u4e0d\u5b58\u5728\u9aa8\u8f6c\u79fb\u75c5\u7076\u4e24\u7c7b\uff09\u9879\u76ee\u5305\u62ec\u4e09\u4e2a\u7814\u7a76\u4efb\u52a1\uff1a\u76ee\u6807\u75c5\u53d8\u533a\u57df\u68c0\u6d4b\uff0c\u7279\u5f81\u6784\u9020\u53ca\u8f6c\u6362\u548c\u5206\u7c7b\u4efb\u52a1\u3002\u5177\u4f53\u5730\uff0c\u91c7\u7528\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u5bf9\u4e0d\u540c\u6a21\u6001\u7684\u533b\u5b66\u56fe\u50cf\u5e8f\u5217\u6570\u636e\u8fdb\u884c\u5355\u72ec\u7684\u9aa8\u8f6c\u79fb\u7624\u76ee\u6807\u68c0\u6d4b\uff0c\u518d\u5bf9\u8fd9\u4e9b\u68c0\u6d4b\u7ed3\u679c\u8fdb\u884c\u7279\u5f81\u63d0\u53d6\u3002\u6240\u63d0\u53d6\u7684\u7279\u5f81\u5305\u62ec\u4e0d\u540c\u6a21\u6001\u4e0b\u68c0\u6d4b\u7ed3\u679c\u7f6e\u4fe1\u5ea6\u7684\u533a\u95f4\u5360\u6bd4\u3001\u68c0\u6d4b\u6846\u7684\u9762\u79ef\u5927\u5c0f\u3001\u68c0\u6d4b\u6846\u5728\u56fe\u50cf\u4e2d\u7684\u7a7a\u95f4\u4f4d\u7f6e\u5206\u5e03\u7b49\u3002\u8fd9\u4e9b\u7279\u5f81\u88ab\u6574\u7406\u6210\u7ed3\u6784\u5316\u6570\u636e\u683c\u5f0f\uff0c\u5b8c\u6210\u4e86\u4ece\u975e\u7ed3\u6784\u5316\u5f71\u50cf\u6570\u636e\u5230\u7ed3\u6784\u5316\u6570\u636e\u7279\u5f81\u7684\u7279\u5f81\u8f6c\u6362\u548c\u878d\u5408\u8fc7\u7a0b\u3002\u6700\u540e\uff0c\u5c06\u8f6c\u6362\u540e\u7684\u7279\u5f81\u8f93\u5165\u5206\u7c7b\u6a21\u578b\u8fdb\u884c\u5206\u7c7b\u4efb\u52a1\u3002\u5b9e\u9a8c\u6bd4\u8f83\u4e86\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u6570\u636e\u878d\u5408\u65b9\u6cd5\u5728\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u80bf\u7624\u5206\u7c7b\u4efb\u52a1\u7684\u6027\u80fd\uff0c\u4e0e\u4ec5\u4f7f\u7528\u5355\u6a21\u6001\u6570\u636e\u6267\u884c\u5206\u7c7b<br \/>\n\u4efb\u52a1\u7684\u6027\u80fd\u3002\u540c\u65f6\uff0c\u8fd8\u5c06\u672c\u9879\u76ee\u63d0\u51fa\u7684\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u878d\u5408\u7b56\u7565\u4e0e\u5176\u4ed6\u878d\u5408\u65b9\u6cd5\u8fdb\u884c\u4e86\u5bf9\u6bd4\u3002<\/p>\n<p><strong>2.2. \u7814\u7a76\u65b9\u6cd5<\/strong><\/p>\n<p><strong>2.2.1. \u7814\u7a76\u8bbe\u8ba1\u548c\u5de5\u4f5c\u6d41\u7a0b<\/strong><\/p>\n<p>\u672c\u9879\u76ee\u65e8\u5728\u5224\u65ad CT\u3001MRI\u3001PET \u56fe\u50cf\u5e8f\u5217\u4e2d\u662f\u5426\u5b58\u5728\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u75c5\u7076\u3002\u5de5\u4f5c\u6d41\u7a0b\u5982\u56fe 2-1 \u6240\u793a\uff1a\u626b\u63cf\u8bbe\u5907\u5bf9\u6bcf\u4f4d\u60a3\u8005\u8fdb\u884c CT\u3001MRI \u6216 PET \u56fe\u50cf\u5e8f\u5217\u7684\u91c7\u96c6\uff1b\u4f7f\u7528\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u5206\u522b\u5728\u4e0d\u540c\u6a21\u6001\u56fe\u50cf\u4e2d\u5bf9\u53ef\u7591\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u7076\u8fdb\u884c\u76ee\u6807\u68c0\u6d4b\uff1b\u5bf9\u68c0\u6d4b\u7ed3\u679c\u8fdb\u884c\u7279\u5f81\u63d0\u53d6\u3001\u6784\u9020\u548c\u878d\u5408\uff0c\u5f97\u5230\u5177\u6709\u53ef\u89e3\u91ca\u6027\u7684\u7ed3\u6784\u5316\u533b\u7597\u6570\u636e\uff1b\u7528\u5206\u7c7b\u6a21\u578b\u5bf9\u7ed3\u6784\u5316\u6570\u636e\u8fdb\u884c\u5206\u7c7b\u4efb\u52a1\uff0c\u5f97\u51fa\u9884\u6d4b\u7ed3\u679c\uff0c\u4ece\u800c\u5224\u65ad\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u662f\u5426\u53d1\u751f\u3002<\/p>\n<p>\u56fe 2-1 \u5de5\u4f5c\u6d41\u7a0b<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-12.14.09\u202fPM.png\" alt=\"img\" \/><\/p>\n<p><strong>2.2.2. \u9aa8\u8f6c\u79fb\u76ee\u6807\u533a\u57df\u68c0\u6d4b<\/strong><\/p>\n<p>\u5148\u7531\u4e24\u4f4d\u4e34\u5e8a\u533b\u5e08\u5bf9\u591a\u6a21\u6001\u6570\u636e\u96c6\u56fe\u50cf\u4e2d\u7684\u9aa8\u8f6c\u79fb\u75c5\u7076\u8fdb\u884c\u4eba\u5de5\u6807\u6ce8\uff0c\u5e76\u5bf9\u60a3\u8005\u8fdb\u884c\u5206\u7c7b\uff08\u6807\u7b7e\u5206\u4e3a\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u548c\u975e\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\uff09\uff0c\u5e76\u8bad\u7ec3 YOLOv5 \u76ee\u6807\u68c0\u6d4b\u7f51\u7edc\uff0c\u4ee5\u8bc6\u522b\u5404\u4e2a\u5355\u4e00\u6a21\u6001\u56fe\u50cf\u4e2d\u7684\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u75c5\u7076\u3002<\/p>\n<p><strong>2.2.3. \u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u6570\u636e\u878d\u5408<\/strong><\/p>\n<p>\u5728\u672c\u9879\u76ee\u7684\u6570\u636e\u96c6\u4e2d\uff0c\u5404\u79cd\u6a21\u6001\u5e8f\u5217\u5f71\u50cf\u7684\u626b\u63cf\u8303\u56f4\u5747\u6db5\u76d6\u4e86\u60a3\u8005\u7684\u5168\u8eab\u3002\u67d0\u60a3\u8005\u7684\u5f71\u50cf\u5e8f\u5217\uff08\u4e0d\u8bba\u662f\u5355\u6a21\u6001\u56fe\u50cf\u8fd8\u662f\u591a\u6a21\u6001\u56fe\u50cf\uff09\u4e2d\u68c0\u6d4b\u5230\u542b\u6709\u9aa8\u8f6c\u79fb\u75c5\u7076\u7684\u5207\u7247\u56fe\u50cf\u6570\u91cf\u8d8a\u591a\uff0c\u5219\u610f\u5473\u7740\u8be5\u60a3\u8005\u53d1\u751f\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u7684\u6982\u7387\u8d8a\u5927\u3002\u6839\u636e\u8fd9\u4e00\u57fa\u672c\u63a8\u7406\uff0c\u91c7\u7528\u540e\u878d\u5408\u65b9\u6cd5\uff0c\u5c06\u4e00\u4e2a\u5f71\u50cf\u5e8f\u5217\u4e2d\u542b\u6709\u80bf\u7624\u5207\u7247\u56fe\u50cf\u7684\u6bd4\u4f8b\uff08\u767e\u5206\u6bd4\uff09\u4f5c\u4e3a\u7ed3\u6784\u5316\u7684\u6570\u636e\u7279\u5f81\uff0c\u4f5c\u4e3a\u540e\u7eed\u5206\u7c7b\u4efb\u52a1\u7684\u4f9d\u636e\u3002<\/p>\n<p>\u5177\u4f53\u64cd\u4f5c\u5982\u4e0b\uff1a\u5728\u6bcf\u4e2a\u6a21\u6001\u7684\u56fe\u50cf\u4e2d\u5b8c\u6210\u9aa8\u8f6c\u79fb\u533a\u57df\u7684\u76ee\u6807\u68c0\u6d4b\u4efb\u52a1\u8bad\u7ec3\u540e\uff0c\u7edf\u8ba1\u6bcf\u4e2a\u56fe\u50cf\u5e8f\u5217\u4e2d\u68c0\u6d4b\u5230\u8f6c\u79fb\u7624\u76ee\u6807\u533a\u57df\u7684\u68c0\u6d4b\u6846\u6570\u91cf\u3002\u6309\u7167\u68c0\u6d4b\u6846\u7684\u7f6e\u4fe1\u5ea6\u5212\u5206\u4e3a 8 \u4e2a\u533a\u95f4\uff1a10%\uff5e20%\u300120%\uff5e30%\u300130%\uff5e40%\u300140%\uff5e50%\u300150%\uff5e60%\u300160%\uff5e70%\u300170%\uff5e80%\u548c\u5927\u4e8e 80%\u3002\u5728\u6bcf\u4e2a\u533a\u95f4\u5185\uff0c\u5206\u522b\u7edf\u8ba1\u5404\u6a21\u6001\u56fe\u50cf\u5e8f\u5217\u4e2d\u8f6c\u79fb\u7624\u68c0\u6d4b\u6846\u6570\u91cf\uff0c\u518d\u9664\u4ee5\u8be5\u5e8f\u5217\u4e2d\u5207\u7247\u56fe\u50cf\u7684\u603b\u6570\uff0c\u5f97\u5230\u6bcf\u4e2a\u7f6e\u4fe1\u5ea6\u533a\u95f4\u5185\u6bcf\u79cd\u6a21\u6001\u56fe\u50cf\u5e8f\u5217\u4e2d\u542b\u6709\u8f6c\u79fb\u7624\u68c0\u6d4b\u6846\u7684\u767e\u5206\u6bd4\u3002\u63a5\u7740\u5c06\u4e09\u79cd\u6a21\u6001\u56fe\u50cf\u63d0\u53d6\u51fa\u7684\u7edf\u8ba1\u7279\u5f81\u62fc\u63a5\uff0c\u7ec4\u6210\u7ed3\u6784\u5316\u6570\u636e\uff0c\u5b9e\u73b0\u8de8\u6a21\u6001\u6570\u636e\u878d\u5408\u3002\u82e5\u60a3\u8005\u7f3a\u5931\u67d0\u79cd\u6a21\u6001\u6570\u636e\uff0c\u76f8\u5e94\u7684\u7edf\u8ba1\u7279\u5f81\uff08\u767e\u5206\u6bd4\uff09\u5c06\u88ab\u7f6e\u4e3a\u96f6\u3002\u7279\u5f81\u8f6c\u6362\u540e\u7684\u7ed3\u6784\u5316\u6570\u636e\u5982\u56fe 2-2 \u6240\u793a\uff0c\u6bcf\u79cd\u6a21\u6001\u6570\u636e\u5305\u62ec 8 \u4e2a\u7279\u5f81\uff0c\u5373\u4e0d\u540c\u7684\u7f6e\u4fe1\u533a\u95f4\uff0c\u6700\u540e\u4e00\u5217\u4e3a\u6807\u7b7e\u503c\uff0c\u5176\u4e2d\u201c0\u201d\u8868\u793a\u8d1f\u4f8b\uff0c\u201c1\u201d\u8868\u793a\u6b63\u4f8b\u3002<\/p>\n<p>\u56fe 2-2 \u7279\u5f81\u8f6c\u6362\u540e\u7684\u7ed3\u6784\u5316\u6570\u636e<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-12.15.10\u202fPM.png\" alt=\"img\" \/><\/p>\n<p><strong>2.2.4. \u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u7684\u5206\u7c7b\u6a21\u578b<\/strong><\/p>\n<p>\u5229\u7528\u6784\u5efa\u597d\u7684\u7ed3\u6784\u5316\u533b\u7597\u7279\u5f81\u8fdb\u884c\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u5206\u7c7b\u4efb\u52a1\uff0c\u878d\u5408\u8de8\u6a21\u6001\u56fe\u50cf\u6570\u636e\u7279\u5f81\u5224\u65ad\u662f\u5426\u53d1\u751f\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u3002\u672c\u9879\u76ee\u91c7\u7528\u5206\u7c7b\u6a21\u578b\u4e3b\u8981\u4ee5\u6a21\u5f0f\u8bc6\u522b\u57fa\u7840\u6a21\u578b\u4e3a\u4e3b\uff0c\u5305\u62ecSVM[59]\u3001AdaBoost[60]\u3001RandomForest[61]\u3001LightGBM[62]\u3001GBDT[63]\u3002SVM \u662f\u4e00\u79cd\u57fa\u4e8e\u6838\u51fd\u6570\u7684\u76d1\u7763\u5b66\u4e60\u6a21\u578b\uff0c\u7528\u4e8e\u89e3\u51b3\u5206\u7c7b\u95ee\u9898\uff0c\u901a\u8fc7\u5bfb\u627e\u6700\u4f18\u8d85\u5e73\u9762\u5728\u7279\u5f81\u7a7a\u95f4\u4e2d\u5c06\u6837\u672c\u5206\u4e3a\u4e0d\u540c\u7c7b\u522b\uff0c\u51b3\u7b56\u51fd\u6570\u6620\u5c04\u8f93\u5165\u7279\u5f81\u5230\u8f93\u51fa\u6807\u7b7e\uff0c\u6838\u51fd\u6570\u5c06\u7279\u5f81\u6620\u5c04\u5230\u65b0\u7a7a\u95f4\uff0c\u635f\u5931\u51fd\u6570\u5ea6\u91cf\u51b3\u7b56\u51fd\u6570\u6027\u80fd\uff0c\u6700\u5927\u5316\u8d85\u5e73\u9762\u4e0e\u6837\u672c\u95f4\u8ddd\u79bb\u5b9e\u73b0\u5206\u7c7b\uff0c\u53ef\u4f7f\u7528\u4e0d\u540c\u6838\u51fd\u6570\u5904\u7406\u9ad8\u7ef4\u7279\u5f81\u3002Adaboost \u662f\u4e00\u79cd\u8fed\u4ee3\u7b97\u6cd5\uff0c\u4e8e1995 \u5e74\u7531 Freund Y \u7b49\u4eba\u63d0\u51fa\uff0c\u80fd\u591f\u5c06\u591a\u4e2a\u5f31\u5206\u7c7b\u5668\u7ed3\u5408\u6210\u4e00\u4e2a\u5f3a\u5206\u7c7b\u5668\uff0c\u901a\u8fc7\u9009\u62e9\u521d\u59cb\u8bad\u7ec3\u96c6\u3001\u8bad\u7ec3\u5f31\u5206\u7c7b\u5668\u3001\u52a0\u6743\u91cd\u65b0\u5206\u914d\u6837\u672c\u548c\u91cd\u590d\u8bad\u7ec3\u76f4\u5230\u8bad\u7ec3\u5b8c\u6210\u6240\u6709\u5f31\u5b66\u4e60\u5668\uff0c\u6700\u540e\u901a\u8fc7\u52a0\u6743\u5e73\u5747\u6216\u6295\u7968\u5f97\u51fa\u6700\u7ec8\u51b3\u7b56\u3002\u7531 Breiman L \u7b49\u4eba\u4e8e 2001 \u5e74\u63d0\u51fa\u5f97 RandomForest \u662f\u4e00\u79cd\u57fa\u4e8e\u51b3\u7b56\u6811\u7684\u673a\u5668\u5b66\u4e60\u7b97\u6cd5\uff0c\u53ef\u7528\u4e8e\u5206\u7c7b\u548c\u56de\u5f52\u4efb\u52a1\u3002\u901a\u8fc7\u6784\u5efa\u591a\u4e2a\u51b3\u7b56\u6811\u5e76\u5bf9\u5b83\u4eec\u7684\u9884\u6d4b\u7ed3\u679c\u8fdb\u884c\u5e73\u5747\u6216\u6295\u7968\u6765\u5f97\u51fa\u6700\u7ec8\u9884\u6d4b\u7ed3\u679c\uff0c\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u968f\u673a\u9009\u62e9\u7279\u5f81\uff0c\u4ee5\u907f\u514d\u8fc7\u62df\u5408\u5e76\u51cf\u5c11\u8ba1\u7b97\u91cf\u3002\u673a\u5668\u5b66\u4e60\u6a21\u578b LightGBM\u662f\u4e00\u79cd\u57fa\u4e8e\u51b3\u7b56\u6811\u7684\u68af\u5ea6\u63d0\u5347\u673a\u7b97\u6cd5\uff0c\u7531Ke G \u7b49\u4eba\u57282017 \u5e74\u63d0\u51fa\uff0c\u9002\u7528\u4e8e\u7ed3\u6784\u5316\u6570\u636e\u7684\u5206\u7c7b\u4efb\u52a1\u3002\u5177\u6709\u9ad8\u6548\u3001\u5185\u5b58\u53cb\u597d\u3001\u652f\u6301\u5e76\u884c\u5904\u7406\u548c\u591a CPU \u7b49\u7279\u70b9\uff0c\u80fd\u5feb\u901f\u5904\u7406\u5927\u91cf\u7279\u5f81\uff0c\u901a\u8fc7\u57fa\u4e8e\u76f4\u65b9\u56fe\u7684\u51b3\u7b56\u6811\u7b97\u6cd5\u51cf\u5c11\u8bad\u7ec3\u65f6\u95f4\u548c\u5185\u5b58\u4f7f\u7528\u91cf\u3002\u901a\u8fc7\u635f\u5931\u51fd\u6570\u7684\u6cf0\u52d2\u5c55\u5f00\u5f0f\u6765\u8fd1\u4f3c\u8868\u793a\u6b8b\u5dee\u6765\u8ba1\u7b97\u635f\u5931\u51fd\u6570\u3002\u7531 Friedman J H \u7b49\u4eba\u4e8e2001 \u5e74\u63d0\u51fa\u7684 GBDT \u662f\u4e00\u79cd\u8fed\u4ee3\u7684\u51b3\u7b56\u6811\u7b97\u6cd5\uff0c\u901a\u8fc7\u6784\u5efa\u591a\u4e2a\u51b3\u7b56\u6811\u6765\u62df\u5408\u76ee\u6807\u51fd\u6570\uff0c\u6bcf\u4e00\u6b65\u90fd\u5728\u4e0a\u4e00\u6b65\u7684\u57fa\u7840\u4e0a\u6784\u5efa\u65b0\u7684\u51b3\u7b56\u6811\uff0c\u4ee5\u4e0d\u65ad\u51cf\u5c0f\u8bef\u5dee\uff0c\u6d41\u7a0b\u5305\u62ec\u9009\u53d6\u5b50\u96c6\u3001\u8bad\u7ec3\u5f31\u5b66\u4e60\u5668\u3001\u68af\u5ea6\u4e0b\u964d\u6cd5\u6700\u5c0f\u5316\u8bef\u5dee\uff0c\u6700\u7ec8\u5c06\u5f31\u5b66\u4e60\u5668\u52a0\u5165\u603b\u4f53\u6a21\u578b\uff0c\u91cd\u590d\u4ee5\u4e0a\u6b65\u9aa4\u76f4\u81f3\u8fbe\u5230\u6700\u4f18\u89e3\u3002<\/p>\n<p><strong>2.2.4.1. \u57fa\u4e8eC3D \u7684\u8de8\u6a21\u6001\u6570\u636e\u878d\u5408\u5206\u7c7b\u6a21\u578b<\/strong><\/p>\n<p>\u672c\u9879\u76ee\u91c7\u7528C3D[64 \u5206\u7c7b\u6a21\u578b\u4f5c\u4e3a\u5bf9\u7167\u6a21\u578b\uff0c\u57fa\u4e8e3D \u5377\u79ef\u795e\u7ecf\u7f51\u7edc\u7684\u6df1\u5ea6\u5b66\u4e60\u65b9\u6cd5\uff0c\u4f7f\u7528\u8de8\u6a21\u6001\u6570\u636e\u878d\u5408\u4e2d\u7684\u524d\u878d\u5408\u7b56\u7565\u3002\u5982\u56fe 2-1 \u6240\u793a\uff0c\u8be5\u878d\u5408\u7b56\u7565\u4ece\u6bcf\u4e2a\u6a21\u6001\u7684\u56fe\u50cf\u5e8f\u5217\u4e2d\u7b5b\u9009\u51fa\u4e00\u90e8\u5206\uff0c\u5408\u5e76\u4e00\u4e2a\u5b8c\u6574\u7684\u591a\u6a21\u6001\u56fe\u50cf\u5e8f\u5217\uff0c\u5e76\u5728\u901a\u9053\u4e0a\u8fdb\u884c\u7ea7\u8054\uff0c\u8fdb\u884c\u8de8\u6a21\u6001\u6570\u636e\u878d\u5408\u3002\u878d\u5408\u540e\u7684\u6570\u636e\u4f5c\u4e3a 3D \u5377\u79ef\u795e\u7ecf\u7f51\u7edc\u7684\u8f93\u5165\uff0c\u7ecf\u8fc7\u591a\u4e2a 3D \u5377\u79ef\u5c42\u63d0\u53d6\u7279\u5f81\uff0c\u6700\u7ec8\u5728\u5168\u8fde\u63a5\u5c42\u4e2d\u6267\u884c\u5206\u7c7b\u4efb\u52a1\uff0c\u4ee5\u5224\u65ad\u5f71\u50cf\u4e2d\u662f\u5426\u5b58\u5728\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u75c5\u7076\u3002<\/p>\n<p><strong>2.3. \u5b9e\u9a8c<\/strong><\/p>\n<p><strong>2.3.1. \u4e34\u5e8a\u4fe1\u606f\u548c\u5f71\u50cf\u6570\u636e\u96c6<\/strong><\/p>\n<p>\u672c\u9879\u76ee\u9009\u53d6\u7b26\u5408\u4ee5\u4e0b\u6761\u4ef6\u7684\u60a3\u8005\u5f00\u5c55\u7814\u7a76\uff1a1\uff09\u4e8e 2000 \u5e74 01 \u6708\u81f3 2020 \u5e74 12\u6708\u671f\u95f4\u5728\u5317\u4eac\u534f\u548c\u533b\u9662\u6216\u56fd\u5bb6\u764c\u75c7\u4e2d\u5fc3\u4e2d\u56fd\u533b\u5b66\u79d1\u5b66\u9662\u80bf\u7624\u533b\u9662\u88ab\u8bca\u65ad\u4e3a\u539f\u53d1\u6027\u4e73\u817a\u764c\u7684\u60a3\u8005 2\uff09\u6709\u5168\u8eab CT \u6216 PET \u6216 MRI \u5176\u4e2d\u4efb\u4e00\u6a21\u6001\u7684\u5168\u8eab\u5f71\u50cf\u6570\u636e\uff1b3\uff09\u6709\u7535\u5b50\u75c5\u5386\u8bb0\u5f55\u3002\u5165\u7ec4\u60a3\u8005\u4e2d\u6709145\u540d\u88ab\u786e\u8bca\u4e3a\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\uff0c\u4f5c\u4e3a\u6b63\u4f8b\u6837\u672c\uff0c\u670988\u540d\u60a3\u8005\u672a\u53d1\u751f\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\uff0c\u4f5c\u4e3a\u8d1f\u4f8b\u6837\u672c\u3002\u6bcf\u4f8b\u6837\u672c\u6570\u636e\u5305\u542b\u4e00\u81f3\u4e09\u79cd\u4e0d\u540c\u6a21\u6001\u7684\u56fe\u50cf\u5e8f\u5217\uff0c\u5176\u56fe\u50cf\u5c3a\u5bf8\u548c\u5207\u7247\u56fe\u50cf\u6570\u91cf\u5404\u5f02\u3002\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u7684\u591a\u6a21\u6001\u533b\u5b66\u56fe\u50cf\u6570\u636e\u96c6\u5bf9\u60a3\u8005\u7684\u5168\u8eab\u8fdb\u884c\u91c7\u6837\uff0c\u7531\u4e8e\u60a3\u8005\u7684 CT\u3001MRI \u6216 PET \u662f\u4e0d\u540c\u65f6\u95f4\u3001\u5728\u4e0d\u540c\u8bbe\u5907\u4e0a\u91c7\u96c6\u7684\uff0c\u4e0d\u540c\u6a21\u6001\u95f4\u7684\u7279\u5f81\u5e76\u975e\u4e00\u4e00\u5339\u914d\u3002\u5176\u4e2d\uff0cCT \u6a21\u6001\u5171\u67093051 \u5f20\u5207\u7247\uff0cMRI \u6a21\u6001\u5171\u6709 3543 \u5f20\u5207\u7247\uff0c\u800c PET \u6a21\u6001\u5171\u6709 1818 \u5f20\u5207\u7247\u3002\u5728\u5165\u7ec4\u8fdb\u884c\u5206\u6790\u4e4b\u524d\uff0c\u6240\u6709\u60a3\u8005\u6570\u636e\u90fd\u5df2\u53bb\u6807\u8bc6\u5316\u3002\u672c\u7814\u7a76\u5df2\u901a\u8fc7\u5317\u4eac\u534f\u548c\u533b\u9662\u4f26\u7406\u59d4\u5458\u4f1a\u6279\u51c6\u3002<\/p>\n<p>\u8be5\u6570\u636e\u96c6\u53ef\u4ee5\u7528\u4e8e\u6267\u884c\u76ee\u6807\u68c0\u6d4b\u4efb\u52a1\u548c\u5206\u7c7b\u4efb\u52a1\u3002<\/p>\n<p>\u9aa8\u8f6c\u79fb\u76ee\u6807\u68c0\u6d4b\u4efb\u52a1\u4ec5\u5206\u6790\u6570\u636e\u96c6\u4e2d\u7684\u6b63\u4f8b\u6837\u672c\uff0c\u8fdb\u884c\u4e94\u6298\u4ea4\u53c9\u9a8c\u8bc1\uff1a\u5c06 145\u4f8b\u60a3\u8005\u7684\u6570\u636e\u6309\u6a21\u6001\u5206\u4e3a\u4e09\u7ec4\uff08CT\u7ec4\u3001MRI\u7ec4\u3001PET\u7ec4\uff09\uff0c\u5728\u6bcf\u4e2a\u7ec4\u5185\u5bf9\u6570\u636e\u8fdb\u884c\u4e94\u7b49\u5206\uff0c\u5728\u6bcf\u8f6e\u9a8c\u8bc1\u4e2d\u9009\u53d6\u4e00\u90e8\u5206\u4f5c\u4e3a\u6d4b\u8bd5\u96c6\u3002\u4e3a\u83b7\u5f97\u6d4b\u8bd5\u6027\u80fd\u7684\u5e73\u5747\u503c\uff0c\u6240\u6709\u6a21\u578b\u90fd\u9700\u8fdb\u884c5 \u6b21\u8bad\u7ec3\u548c\u8bc4\u4f30\u3002<\/p>\n<p>\u5728\u5229\u7528\u7ed3\u6784\u5316\u6570\u636e\u6267\u884c\u5206\u7c7b\u4efb\u52a1\u65f6\uff0c\u9700\u8981\u5e73\u8861\u6b63\u8d1f\u6837\u672c\u6570\u91cf\uff0c\u56e0\u6b64\u8981\u6269\u5145\u6570\u636e\u96c6\u3002\u5c06\u5177\u6709\u591a\u79cd\u6a21\u6001\u7684\u6837\u672c\u62c6\u5206\u4e3a\u5305\u542b\u8f83\u5c11\u6a21\u6001\u7684\u6837\u672c\uff0c\u5982\u5c06\u201cCT+MRI+PET\u201d\u7c7b\u578b\u62c6\u5206\u4e3a\u201cCT+MRI\u201d\u6216\u201cCT+PET\u201d\u7b49\u3002\u5982\u8868 2-1 \u6240\u793a\uff0c\u6269\u5145\u540e\u5171\u6709 380 \u4f8b\u6837\u672c\u6570\u636e\uff0c\u5305\u62ec188 \u4e2a\u6b63\u6837\u672c\u548c 192 \u4e2a\u8d1f\u6837\u672c\u3002\u4e0b\u4e00\u6b65\uff0c\u5408\u5e76\u4e94\u6298\u4ea4\u53c9\u9a8c\u8bc1\u7684\u76ee\u6807\u68c0\u6d4b\u7ed3\u679c\uff0c\u6b64\u540e\uff0c\u8fdb\u884c\u7279\u5f81\u6784\u5efa\u548c\u8f6c\u6362\uff0c\u4ece\u800c\u83b7\u5f97\u9002\u5408\u8de8\u6a21\u6001\u6570\u636e\u878d\u5408\u548c\u5206\u7c7b\u4efb\u52a1\u7684\u7ed3\u6784\u5316\u6570\u636e\uff1b\u5bf9\u4e8e\u8d1f\u6837\u672c\u6570\u636e\uff0c\u4e5f\u9700\u8981\u5728\u5408\u5e76\u9aa8\u8f6c\u79fb\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u7684\u63a8\u7406\u7ed3\u679c\u540e\uff0c\u5bf9\u6570\u636e\u8fdb\u884c\u7ed3\u6784\u5316\u5904\u7406\u3002<\/p>\n<p>\u4e3a\u8bc1\u5b9e\u5728\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u5224\u65ad\u7684\u5206\u7c7b\u4efb\u52a1\u4e2d\uff0c\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u878d\u5408\u6570\u636e\u6027\u80fd\u4f18\u4e8e\u5355\u4e00\u6a21\u6001\u6570\u636e\uff0c\u9700\u8981\u8fdb\u884c\u591a\u6a21\u6001\u878d\u5408\u6570\u636e\u4e0e\u5355\u6a21\u6001\u6570\u636e\u7684\u5bf9\u7167\u5b9e\u9a8c\u3002\u5982\u8868 2-1 \u6240\u793a\uff0c\u5355\u6a21\u6001\u6570\u636e\u5305\u62ec\u4ec5\u6709 CT\u3001\u4ec5\u6709 MRI \u548c\u4ec5\u6709 PET \u4e09\u79cd\u7c7b\u578b\u7684\u6570\u636e\u96c6\u5408\uff0c\u603b\u8ba1 212 \u4e2a\u6837\u672c\uff0c\u800c\u591a\u6a21\u6001\u6570\u636e\u6db5\u76d6\u4e86CT+MRI\u3001CT+PET\u3001MRI+PET \u548cCT+MRI+PET \u56db\u79cd\u7c7b\u578b\uff0c\u5171\u8ba1 168 \u4e2a\u6837\u672c\u3002\u5206\u522b\u5bf9\u5355\u6a21\u6001\u6570\u636e\u548c\u591a\u6a21\u6001\u6570\u636e\u8fdb\u884c\u72ec\u7acb\u5212\u5206\uff0c\u5c06\u6bcf\u79cd\u6a21\u6001\u6570\u636e\u8fdb\u884c\u4e94\u7b49\u4efd\uff0c\u8fdb\u884c\u4e94\u6298\u4ea4\u53c9\u9a8c\u8bc1\u3002\u5728\u6bcf\u8f6e\u9a8c\u8bc1\u4e2d\uff0c\u9009\u62e9\u4e00\u90e8\u5206\u4f5c\u4e3a\u6d4b\u8bd5\u96c6\u3002\u5229\u7528 SVM\u3001AdaBoost\u3001RandomForest\u3001LightGBM\u3001GBDT\u4ee5\u53ca C3D \u6a21\u578b\u8fdb\u884c\u5b9e\u9a8c\uff0c\u6bcf\u4e2a\u6a21\u578b\u90fd\u9700\u8fdb\u884c 5 \u8f6e\u8bad\u7ec3\u548c\u8bc4\u4f30\uff0c\u4ee5\u83b7\u5f97\u6d4b\u8bd5\u96c6\u4e0a\u6027\u80fd\u7684\u5e73\u5747\u503c\u3002<\/p>\n<p>\u4e3a\u9002\u5e94 C3D \u6a21\u578b\u5bf9\u56fe\u50cf\u7edf\u4e00\u5c3a\u5bf8\u7684\u8981\u6c42\uff0c\u9488\u5bf9\u4e0d\u540c\u60a3\u8005\u5207\u7247\u6570\u91cf\u3001\u5927\u5c0f\u7684\u5dee\u5f02\uff0c\u8fdb\u884c\u9884\u5904\u7406\u3002\u5728\u6bcf\u79cd\u6a21\u6001\u56fe\u50cf\u5e8f\u5217\u4e2d\u7b49\u95f4\u9694\u62bd\u53d6 60 \u5f20\u56fe\u50cf\u5207\u7247\uff0c\u5e76\u8fdb\u884c\u7f29\u653e\uff0c\u4f7f\u5176\u7ec4\u5408\u4e3a 180 \u5f20 128\u00d7128 \u50cf\u7d20\u7684\u5207\u7247\u3002\u5bf9\u4e8e\u7f3a\u5931\u7684\u6a21\u6001\u6570\u636e\uff0c\u7528 60 \u5f20\u96f6\u50cf\u7d20\u503c\u7684\u9ed1\u8272\u56fe\u50cf\u5207\u7247\u8fdb\u884c\u586b\u5145\u3002\u4ece 180 \u5f20\u5207\u7247\u4e2d\u968f\u673a\u9009\u53d6\u4e00\u4e2a\u8d77\u59cb\u4f4d\u7f6e\uff0c\u8fde\u7eed\u62bd\u53d6 120\u5f20\u5207\u7247\u4f5c\u4e3a\u6a21\u578b\u7684\u6700\u7ec8\u8f93\u5165\uff0c\u786e\u4fdd\u8f93\u5165\u5c3a\u5bf8\u4e3a 128\u00d7128\u00d7120 \u50cf\u7d20\u3002<\/p>\n<p>\u8868 2-1 \u6269\u5145\u540e\u7684\u5206\u7c7b\u6570\u636e\u96c6<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-12.18.14\u202fPM.png\" alt=\"img\" \/><\/p>\n<p><strong>2.3.2. \u6a21\u578b\u8bad\u7ec3\u8fc7\u7a0b<\/strong><\/p>\n<p>\u5728\u6309\u4e0a\u8ff0\u6b65\u9aa4\u51c6\u5907\u597d\u6570\u636e\u96c6\u540e\uff0c\u8fdb\u884c\u76ee\u6807\u68c0\u6d4b\u4efb\u52a1\u8bad\u7ec3\u65f6\uff0c\u5c06\u6bcf\u4e2a\u6a21\u6001\u7684\u56fe\u50cf\u5927\u5c0f\u7edf\u4e00\u52301024\u00d71024\u50cf\u7d20\uff0c\u7136\u540e\u4f7f\u7528\u591a\u79cd\u6570\u636e\u589e\u5f3a\u65b9\u6cd5\uff0c\u589e\u52a0\u8f93\u5165\u6570\u636e\u96c6\u5bf9\u566a\u58f0\u7684\u9c81\u68d2\u6027\u3002<\/p>\n<p>\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u91c7\u7528 YOLOv5 \u6a21\u578b\u6a21\u578b\u5e76\u4f7f\u7528 PyTorch \u6df1\u5ea6\u5b66\u4e60\u6846\u67b6\u5728 2 \u4e2aNvidia Tesla V100-SXM2 32G GPU \u4e0a\u8fdb\u884c 70 \u4e2a\u8f6e\u6b21\u7684\u8bad\u7ec3\u3002\u521d\u59cb\u5b66\u4e60\u7387\u4e3a 0.00001\uff0c\u4f7f\u7528 0.98 \u7684\u52a8\u91cf\u548c 0.01 \u7684\u6743\u503c\u8870\u51cf\u901a\u8fc7 SGD \u6765\u4f18\u5316\u5404\u7f51\u7edc\u5c42\u7684\u6743\u91cd\u76ee\u6807\u51fd\u6570\u3002\u5728\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\uff0c\u7f51\u7edc\u5728\u9a8c\u8bc1\u96c6\u4e0a\u8fbe\u5230\u6700\u5c0f\u7684\u635f\u5931\u65f6\uff0c\u9009\u62e9\u6700\u4f73\u53c2\u6570\u3002<\/p>\n<p>\u8fdb\u884c\u5206\u7c7b\u4efb\u52a1\u65f6\uff0c\u91c7\u7528 LightGBM\u3001GBDT\u3001AdaBoost\u3001RandomForest \u4ee5\u53caSVM\uff0c\u4e0a\u8ff0\u6a21\u578b\u5747\u4e3a\u673a\u5668\u5b66\u4e60\u6a21\u578b\uff0c\u5176\u8d85\u53c2\u6570\u5bf9\u6a21\u578b\u9884\u6d4b\u7ed3\u679c\u5f71\u54cd\u8f83\u5927\uff0c\u5728\u5206\u7c7b\u4efb\u52a1\u4e2d\uff0c\u91c7\u7528SVM\u3001AdaBoost\u3001RandomForest\u3001LightGBM\u4ee5\u53caGBDT\u7b49\u673a\u5668\u5b66\u4e60\u6a21\u578b\u3002\u56e0\u5176\u8d85\u53c2\u6570\u4f1a\u5bf9\u9884\u6d4b\u7ed3\u679c\u4ea7\u751f\u8f83\u5927\u5f71\u54cd\uff0c\u5728\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\uff0c\u4f7f\u7528\u7f51\u683c\u641c\u7d22\u7b56\u7565\u4e3a\u8fd9\u4e9b\u6a21\u578b\u5bfb\u627e\u6700\u4f73\u53c2\u6570\u3002\u7f51\u683c\u641c\u7d22\u7b56\u7565\u5728\u4e00\u5b9a\u8303\u56f4\u7684\u8d85\u53c2\u6570\u7a7a\u95f4\u5185\u5bfb\u627e\u6700\u4f73\u7684\u8d85\u53c2\u6570\u7ec4\u5408\uff0c\u901a\u8fc7\u679a\u4e3e\u5404\u79cd\u53ef\u80fd\u7684\u7ec4\u5408\u5e76\u8bc4\u4f30\u6a21\u578b\u9884\u6d4b\u7ed3\u679c\uff0c\u6700\u7ec8\u9009\u62e9\u8868\u73b0\u6700\u4f18\u7684\u8d85\u53c2\u6570\u7ec4\u5408\u3002\u8981\u641c\u7d22\u7684\u8d85\u53c2\u6570\u5305\u62ec\u5b66\u4e60\u7387\u3001\u6811\u7684\u6700\u5927\u6df1\u5ea6\u3001\u53f6\u5b50\u8282\u70b9\u6570\u91cf\u3001\u968f\u673a\u62bd\u6837\u6bd4\u4f8b\u3001\u6743\u91cd\u7684L1\u6b63\u5219\u5316\u9879\u548c\u6743\u91cd\u7684L2\u6b63\u5219\u5316\u9879\u7b49\u3002\u5b9e\u9a8c\u7ed3\u679c\u5c06\u57fa\u4e8e\u6700\u4f18\u8d85\u53c2\u6570\u8bbe\u5b9a\u4e0b\u7684\u9884\u6d4b\u6a21\u578b\u3002\u6a21\u578b\u8bad\u7ec3\u7684\u7f51\u7edc\u7ed3\u6784\u56fe\u5982\u56fe 2-3 \u6240\u793a\u3002<\/p>\n<p>\u7528\u4e8e\u5bf9\u7167\u7684 C3D \u6a21\u578b\u4f7f\u7528 PyTorch \u6df1\u5ea6\u5b66\u4e60\u6846\u67b6\u5728 1 \u4e2a NVIDIA Tesla V100-SXM2 32GB GPU \u4e0a\u8bad\u7ec3 100 \u4e2a\u8f6e\u6b21\uff0c\u521d\u59cb\u5b66\u4e60\u7387\u4e3a 0.00001\uff0c\u4f7f\u7528\u52a8\u91cf\u4e3a 0.9\uff0c\u6743\u503c\u8870\u51cf\u4e3a0.0005 \u7684 SGD \u68af\u5ea6\u4e0b\u964d\u4f18\u5316\u5668\u5bf9\u5404\u7f51\u7edc\u5c42\u6743\u91cd\u7684\u76ee\u6807\u51fd\u6570\u8fdb\u884c\u4f18\u5316\u3002<\/p>\n<p>\u56fe 2-3 \u7f51\u7edc\u7ed3\u6784\u56fe<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-12.20.01\u202fPM.png\" alt=\"img\" \/><\/p>\n<p><strong>2.3.3. \u8bc4\u4ef7\u6307\u6807<\/strong><\/p>\n<p>\u672c\u9879\u76ee\u4e2d\u7684\u9aa8\u8f6c\u79fb\u76ee\u6807\u68c0\u6d4b\u4efb\u52a1\u91c7\u7528 AP50 \u4f5c\u4e3a\u8bc4\u4ef7\u6307\u6807\uff0c\u5176\u4ecb\u7ecd\u8be6\u89c1\u4e0a\u4e00\u7ae0\u8282\u3002<\/p>\n<p>\u800c\u5728\u5206\u7c7b\u4efb\u52a1\u4e2d\uff0c\u91c7\u7528\u51c6\u786e\u7387\uff08Accuracy, Acc\uff09\u3001\u654f\u611f\u6027\uff08Sensitivity, Sen\uff09\u3001\u7279\u5f02\u6027\uff08Specificity, Spe\uff09\u3001AUC\uff08Area Under Curve, AUC\uff09\u4f5c\u4e3a\u8bc4\u4ef7\u6307\u6807\uff0c\u5e76\u91c7\u7528ROC \u66f2\u7ebf\u548cPR \u66f2\u7ebf\u5bf9\u6a21\u578b\u8fdb\u884c\u8bc4\u4f30\u3002\u51c6\u786e\u7387\u662f\u6307\u5bf9\u4e8e\u7ed9\u5b9a\u7684\u6d4b\u8bd5\u96c6\uff0c\u5206\u7c7b\u6a21\u578b\u6b63\u786e\u5206\u7c7b\u7684\u6837\u672c\u6570\u5360\u603b\u6837\u672c\u6570\u7684\u6bd4\u4f8b\uff0c\u5982\u516c\u5f0f 5 \u6240\u793a\uff0c\u5176\u4e2d\u771f\u6b63\u4f8b\uff08True Positive, TP\uff09\u8868\u793a\u9884\u6d4b\u4e3a\u6b63\u4f8b\u4e14\u6807\u7b7e\u503c\u4e3a\u6b63\u4f8b\uff0c\u5047\u6b63\u4f8b\uff08False Positive, FP\uff09\u8868\u793a\u9884\u6d4b\u4e3a\u6b63\u4f8b\u4f46\u6807\u7b7e\u503c\u4e3a\u8d1f\u4f8b\uff0c\u548c\u5047\u8d1f\u4f8b\uff08False Negative, FN\uff09\u548c\u771f\u8d1f\u4f8b\uff08True Negative, TN\uff09\u4ee3\u8868\u7684\u6982\u5ff5\u4ee5\u6b64\u7c7b\u63a8\u3002\u5982\u516c\u5f0f 64 \u548c\u516c\u5f0f 7 \u6240\u793a\uff0c\u654f\u611f\u6027\u548c\u7279\u5f02\u6027\u7684\u5b9a\u4e49\u5206\u522b\u4e3a\uff1a\u9884<br \/>\n\u6d4b\u6b63\u786e\u7684\u6b63\u4f8b\u5360\u6240\u6709\u6b63\u4f8b\u7684\u6bd4\u4f8b\uff0c\u4ee5\u53ca\u9884\u6d4b\u6b63\u786e\u7684\u8d1f\u4f8b\u5360\u6240\u6709\u8d1f\u4f8b\u7684\u6bd4\u4f8b\u3002ROC\u66f2\u7ebf\u662f\u4e00\u79cd\u8bc4\u4f30\u4e8c\u5206\u7c7b\u6a21\u578b\u7684\u65b9\u6cd5\uff0c\u5176\u6a2a\u8f74\u4e3a\u5047\u9633\u6027\u7387\uff08False Positive Rate\uff0cFPR\uff09\uff0c\u5176\u8ba1\u7b97\u65b9\u5f0f\u4e0e\u4e0a\u4e00\u7ae0\u7684\u53ec\u56de\u7387\uff08Recall\uff09\u76f8\u540c\uff0c\u7eb5\u8f74\u4e3a\u771f\u9633\u6027\u7387\uff08True Positive Rate\uff0cTPR\uff09\uff0cTPR \u548c FPR \u7684\u8ba1\u7b97\u65b9\u5f0f\u8be6\u89c1\u516c\u5f0f 8 \u548c\u516c\u5f0f 9\u3002ROC \u66f2\u7ebf\u5c55\u793a\u4e86\u5728\u4e0d\u540c\u9608\u503c\u4e0b\uff0cTPR \u4e0e FPR \u7684\u53d8\u5316\u5173\u7cfb\u3002\u56e0\u4e3a\u5de6\u4e0a\u89d2\u70b9\u5bf9\u5e94\u7684\u5047\u9633\u6027\u7387\u4e3a 0\uff0c\u771f\u9633\u6027\u7387\u4e3a 1\uff0c\u8868\u660e\u6a21\u578b\u5c06\u6240\u6709\u6b63\u4f8b\u6837\u672c\u5206\u7c7b\u6b63\u786e\uff0c\u4e14\u672a\u5c06\u4efb\u4f55\u8d1f\u4f8b\u6837\u672c\u8bef\u5224\u4e3a\u6b63\u4f8b\u3002\u82e5 ROC \u66f2\u7ebf\u9760\u8fd1\u5de6\u4e0a\u89d2\uff0c\u63d0\u793a\u6a21\u578b\u6027\u80fd\u8f83\u597d\u3002AUC\u4ee3\u8868ROC\u66f2\u7ebf\u4e0b\u7684\u9762\u79ef\uff0c\u5373\u4ece\uff080,0\uff09\u5230\uff081,1\uff09\u8fdb\u884c\u79ef\u5206\u6d4b\u91cfROC\u66f2\u7ebf\u4e0b\u4e8c\u7ef4\u533a\u57df\u7684\u9762\u79ef\u3002AUC\u7efc\u5408\u8003\u8651\u6240\u6709\u53ef\u80fd\u7684\u5206\u7c7b\u9608\u503c\uff0c\u63d0\u4f9b\u4e86\u4e00\u4e2a\u5168\u9762\u7684\u6027\u80fd\u5ea6\u91cf\u3002AUC \u503c\u8868\u793a\u968f\u673a\u4ece\u6b63\u8d1f\u6837\u672c\u4e2d\u5404\u62bd\u53d6\u4e00\u4e2a\uff0c\u5206\u7c7b\u5668\u6b63\u786e\u9884\u6d4b\u6b63\u4f8b\u5f97\u5206\u9ad8\u4e8e\u8d1f\u4f8b\u7684\u6982\u7387\u3002AUC \u503c\u8d8a\u63a5\u8fd1 1\uff0c\u8bf4\u660e\u6a21\u578b\u6027\u80fd\u8d8a\u4f18\u79c0\u3002PR\u66f2\u7ebf\u7684\u7ed8\u5236\u65b9\u6cd5\u8be6\u89c1\u4e0a\u4e00\u7ae0\uff0c PR\u66f2\u7ebf\u5728\u4e0d\u540c\u5206\u7c7b\u9608\u503c\u4e0b\u5c55\u793a\u4e86\u5206\u7c7b\u5668\u5728\u7cbe\u5ea6\uff08Precision, P\uff09\u548c\u53ec\u56de\u7387\uff08Recall, R\uff09\u65b9\u9762\u7684\u6574\u4f53\u8868\u73b0\u3002<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-12.21.04\u202fPM.png\" alt=\"img\" \/><\/p>\n<p><strong>2.3.4. \u5355\u6a21\u6001\u9aa8\u8f6c\u79fb\u7076\u68c0\u6d4b\u6a21\u578b\u53ca\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u5206\u7c7b\u6a21\u578b\u7684\u7ed3\u679c\u4e0e\u5206\u6790<\/strong><\/p>\n<p>\u672c\u9879\u76ee\u5bf9\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u591a\u6a21\u6001\u533b\u5b66\u56fe\u50cf\u6570\u636e\u96c6\uff08\u5305\u62ec CT\u3001MRI\u3001PET\uff09\u8fdb\u884c\u4e86\u5355\u6a21\u6001\u80bf\u7624\u68c0\u6d4b\u5b9e\u9a8c\u548c\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u75c5\u4f8b\u5206\u7c7b\u5b9e\u9a8c\u3002\u5176\u4e2d\uff0c\u5355\u6a21\u6001\u80bf\u7624\u68c0\u6d4b\u5b9e\u9a8c\u662f\u591a\u6a21\u6001\u80bf\u7624\u5206\u7c7b\u5b9e\u9a8c\u7684\u524d\u7f6e\u6b65\u9aa4\u3002<\/p>\n<p>\u91c7\u7528\u5355\u9636\u6bb5\u76ee\u6807\u68c0\u6d4b\u6a21\u578b YOLOv5\uff0c\u5728\u4e94\u6298\u4ea4\u53c9\u9a8c\u8bc1\u4e0b\u6bd4\u8f83\u4e86\u4f7f\u7528\u5355\u6a21\u6001 CT\u56fe\u50cf\u3001PET\u56fe\u50cf\u3001MRI\u56fe\u50cf\u4f5c\u4e3a\u8f93\u5165\u6570\u636e\u96c6\u65f6\uff0c\u6a21\u578b\u7684\u76ee\u6807\u68c0\u6d4b\u6027\u80fd\u3002\u5e76\u5728\u5c06\u76ee\u6807\u68c0\u6d4b\u7ed3\u679c\u8fdb\u884c\u7279\u5f81\u8f6c\u6362\u540e\uff0c\u91c7\u7528\u4e0d\u540c\u7684\u5206\u7c7b\u6a21\u578b\uff0c\u5305\u62ec\u540e\u878d\u5408\u5206\u7c7b\u6a21\u578b\uff08LightGBM\u3001GBDT\u3001AdaBoost\u3001RandomForest\u3001SVM\uff09 \u548c \u524d \u878d \u5408 \u5206 \u7c7b \u6a21 \u578b\uff08C3D\uff09\uff0c\u5728\u4e94\u6298\u4ea4\u53c9\u9a8c\u8bc1\u4e0b\u6bd4\u8f83\u4e86\u4f7f\u7528\u5355\u6a21\u6001\u6570\u636e\u548c\u8de8\u6a21\u6001\u878d\u5408\u6570\u636e\u4f5c\u4e3a\u8f93\u5165\u65f6\uff0c\u6bcf\u4e2a\u6a21\u578b\u7684\u5206\u7c7b\u6027\u80fd\u3002<\/p>\n<p>\u8868 2-2 \u5c55\u793a\u4e86\u5728\u4e0d\u540c\u5355\u4e00\u6a21\u6001\u6570\u636e\u4e0a\uff0c\u4e94\u6298\u4ea4\u53c9\u9a8c\u8bc1\u5f97\u5230\u7684\u9aa8\u8f6c\u79fb\u75c5\u7076\u68c0\u6d4b\u7ed3\u679c\uff0c\u8bc4\u4f30\u6307\u6807\u4e3a AP50\u3002\u5b9e\u9a8c\u7ed3\u679c\u8868\u660e\uff0cPET \u6a21\u6001\u7684\u68c0\u6d4b\u7cbe\u5ea6\u8f83\u9ad8\uff0c\u800c CT \u6a21\u6001\u7684\u68c0\u6d4b\u7cbe\u5ea6\u6700\u4f4e\u3002\u8f93\u5165\u6570\u636e\u91cf\u8f83\u5c11\u3001\u68c0\u6d4b\u76ee\u6807\u9762\u79ef\u5c0f\u3001\u8f6c\u79fb\u7624\u7684\u7279\u5f81\u96be\u4ee5\u4e0e\u6b63\u5e38\u9aa8\u7ec4\u7ec7\u533a\u5206\u662f\u63d0\u9ad8\u68c0\u6d4b\u7cbe\u5ea6\u7684\u96be\u70b9\u3002\u56fe 2-4 \u4e0d\u540c\u5355\u4e00\u6a21\u6001\u56fe\u50cf\u4e0b\u76ee\u6807\u9aa8\u8f6c\u79fb\u75c5\u53d8\u533a\u57df\u7684\u68c0\u6d4b\u7ed3\u679c\u548c\u5b9e\u9645\u6807\u6ce8\u7684\u764c\u7076\u533a\u57df\u8fdb\u884c\u4e86\u53ef\u89c6\u5316\u3002\u7eff\u8272\u6846\u7531\u533b\u5e08\u6807\u6ce8\uff0c\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u6807\u6ce8\u7684\u9884\u6d4b\u6846\u4e3a\u9ec4\u8272\u3002<\/p>\n<p>\u8868 2-2 \u5355\u6a21\u6001\u9aa8\u8f6c\u79fb\u7076\u68c0\u6d4b\u4e94\u6298\u4ea4\u53c9\u9a8c\u8bc1\u7ed3\u679c<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-12.22.14\u202fPM.png\" alt=\"img\" \/><\/p>\n<p>\u56fe 2-4 \u53ef\u89c6\u5316\u5355\u6a21\u6001\u76ee\u6807\u68c0\u6d4b\u7ed3\u679c<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-12.22.39\u202fPM.png\" alt=\"img\" \/><\/p>\n<p>\u5c06 CT\u3001MRI \u548c PET \u7684\u6570\u636e\u7ec4\u6210\u5355\u6a21\u6001\u5b50\u6570\u636e\u96c6\u8fdb\u884c\u5355\u6a21\u6001\u5206\u6790\uff0c\u800c\u5c06\u4e24\u79cd\u53ca\u4e24\u79cd\u4ee5\u4e0a\u7684\u6570\u636e\u7ec4\u6210\u8de8\u6a21\u6001\u5b50\u6570\u636e\u96c6\u8fdb\u884c\u591a\u6a21\u6001\u5206\u6790\u3002\u8868 2-3 \u548c\u8868 2-4 \u5c55\u793a\u4e86\u5728\u4e0a\u8ff0\u4e24\u79cd\u5b50\u6570\u636e\u96c6\u4e2d\u8fdb\u884c\u4e94\u6298\u4ea4\u53c9\u9a8c\u8bc1\u7684\u7ed3\u679c\uff0c\u5bf9\u6bd4\u4e866 \u79cd\u4e0d\u540c\u6a21\u578b\u6bcf\u4e00\u6298\u7684\u51c6\u786e\u7387\u3001AUC\uff0c\u53ca\u5176\u5e73\u5747\u503c\u3002\u8fd9 6 \u7c7b\u6a21\u578b\u4e2d\u7684\u524d 5 \u79cd\u6a21\u578b\u4f7f\u7528\u540e\u878d\u5408\u7b56\u7565\uff0c\u800c\u4f5c\u4e3a\u5bf9\u7167\u7684C3D \u6a21\u578b\u91c7\u7528\u524d\u878d\u5408\u7b56\u7565\u3002\u5bf9\u6bd4\u8868 2-3 \u548c\u8868 2-4 \u7684\u5b9e\u9a8c\u7ed3\u679c\u53ef\u77e5\uff0c\u5728\u4efb\u4e00\u6a21\u578b\uff08\u5305\u62ec\u524d\u878d\u5408\u6a21\u578b\uff09\u4e2d\uff0c\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u878d\u5408\u6570\u636e\u5728\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u5206\u7c7b\u4efb\u52a1\u4e0a\u76f8\u8f83\u4e8e\u4ec5\u4f7f\u7528\u5355\u6a21\u6001\u6570\u636e\u7684\u6027\u80fd\u6709\u6240\u63d0\u9ad8\uff1a\u5e73\u5747\u51c6\u786e\u7387\u63d0\u9ad8\u4e867.9%\uff1b\u5e73\u5747AUC \u63d0\u9ad8\u4e868.5%\u3002\u5982\u88682-5 \u548c2-6 \u6240\u793a\uff0c\u8de8\u6a21\u6001\u878d\u5408\u65b9\u6cd5\u6bd4\u5355\u6a21\u6001\u65b9\u6cd5\u7684\u5e73\u5747\u654f\u611f\u6027\u63d0\u9ad8\u4e867.6%\uff0c\u5e73\u5747\u7279\u5f02\u6027\u63d0\u9ad8\u4e86 9.4%\u3002<\/p>\n<p>\u8868 2-3 \u57fa\u4e8e\u5355\u6a21\u6001\u5b50\u6570\u636e\u96c6\u8fdb\u884c\u5206\u7c7b\u4efb\u52a1\u7684\u51c6\u786e\u7387\u548c AUC<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-12.23.16\u202fPM.png\" alt=\"img\" \/><\/p>\n<p>\u8868 2-4 \u57fa\u4e8e\u8de8\u6a21\u6001\u5b50\u6570\u636e\u96c6\u884c\u5206\u7c7b\u4efb\u52a1\u7684\u51c6\u786e\u7387\u548c AUC<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-12.23.39\u202fPM.png\" alt=\"img\" \/><\/p>\n<p>\u8868 2-5 \u5355\u6a21\u6001\u6570\u636e\u5206\u7c7b\u7684\u654f\u611f\u6027\u548c\u7279\u5f02\u6027<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-12.24.02\u202fPM.png\" alt=\"img\" \/><\/p>\n<p>\u8868 2-6 \u8de8\u6a21\u6001\u878d\u5408\u6570\u636e\u5206\u7c7b\u7684\u654f\u611f\u6027\u548c\u7279\u5f02\u6027<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-12.24.29\u202fPM.png\" alt=\"img\" \/><\/p>\n<p>\u56fe 2-5 \u548c\u56fe 2-6 \u5206\u522b\u5c55\u793a\u4e86 6 \u4e2a\u6a21\u578b\u5229\u7528\u5355\u6a21\u6001\u6570\u636e\u8fdb\u884c\u5206\u7c7b\u5b9e\u9a8c\u548c\u5229\u7528\u7279\u5f81\u8f6c\u6362\u548c\u878d\u5408\u540e\u7684\u8de8\u6a21\u6001\u6570\u636e\u8fdb\u884c\u5206\u7c7b\u5b9e\u9a8c\u7684 PR \u66f2\u7ebf\u3002\u53ef\u4ee5\u6839\u636e\u66f2\u7ebf\u5f62\u72b6\u548c\u66f2\u7ebf\u4e0b\u65b9\u9762\u79ef\u6765\u8bc4\u4f30\u4e0d\u540c\u6a21\u578b\u7684\u6027\u80fd\u8868\u73b0\uff0c\u66f2\u7ebf\u4e0b\u9762\u79ef\u8d8a\u5927\uff0c\u63d0\u793a\u6a21\u578b\u7684\u6027\u80fd\u8d8a\u4f18\u79c0\u3002\u7efc\u5408\u89c2\u5bdf\u5355\u6a21\u6001\u548c\u8de8\u6a21\u6001\u5206\u7c7b\u5b9e\u9a8c\u7684P-R\u66f2\u7ebf\u56fe\uff0c\u53ef\u4ee5\u53d1\u73b0\uff0c\u57fa\u4e8e\u8de8\u6a21\u6001\u6570\u636e\u7684\u5206\u7c7b\u4efb\u52a1\u7684P-R \u66f2\u7ebf\u4e0b\u9762\u79ef\u5927\u4e8e\u57fa\u4e8e\u5355\u6a21\u6001\u6570\u636e\u7684\u5206\u7c7b\u4efb\u52a1\u7684P-R \u66f2\u7ebf\u4e0b\u9762\u79ef\uff0c\u63d0\u793a\u8de8\u6a21\u6001\u6570\u636e\u4f5c\u4e3a\u8f93\u5165\u65f6\u5206\u7c7b\u6a21\u578b\u7684\u8868\u73b0\u66f4\u52a0\u51fa\u8272\u3002<\/p>\n<p>\u6bd4\u8f83\u57fa\u4e8e\u5355\u6a21\u6001\u6570\u636e\u8fdb\u884c\u5206\u7c7b\u7684\u6a21\u578b\u7684 P-R \u66f2\u7ebf\uff0c\u53ef\u89c1 3D \u5377\u79ef\u7f51\u7edc\u7684\u8bad\u7ec3\u65b9\u5f0f\u76f8\u8f83\u4e8e\u5176\u4ed6\u540e\u878d\u5408\u6a21\u578b\u7684\u6027\u80fd\u8868\u73b0\u66f4\u4f18\u3002\u7136\u800c\uff0c\u5728\u57fa\u4e8e\u8de8\u6a21\u6001\u6570\u636e\u8fdb\u884c\u5206\u7c7b\u7684\u6a21\u578b\u7684 P-R \u66f2\u7ebf\u4e2d\uff0c\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u540e\u878d\u5408\u7b56\u7565\u76f8\u5bf9\u4e8e\u57fa\u4e8e 3D \u5377\u79ef\u7684\u524d\u878d\u5408\u65b9\u6cd5\u5177\u6709\u66f4\u597d\u7684\u6027\u80fd\u3002<\/p>\n<p>\u56fe 2-5 \u57fa\u4e8e\u5355\u6a21\u6001\u6570\u636e\u4e0d\u540c\u5206\u7c7b\u6a21\u578b\u7684\u7684PR \u66f2\u7ebf<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-12.25.55\u202fPM.png\" alt=\"img\" \/><\/p>\n<p>\u56fe 2-6 \u57fa\u4e8e\u8de8\u6a21\u6001\u6570\u636e\u4e0d\u540c\u5206\u7c7b\u6a21\u578b\u7684\u7684PR \u66f2\u7ebf<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-12.26.22\u202fPM.png\" alt=\"img\" \/><\/p>\n<p>\u56fe 2-7 \u548c\u56fe 2-8 \u5c55\u793a\u4e86 6 \u79cd\u5206\u7c7b\u6a21\u578b\u5728\u4f7f\u7528\u5355\u6a21\u6001\u6570\u636e\u8fdb\u884c\u5206\u7c7b\u5b9e\u9a8c\u548c\u4f7f\u7528\u8de8\u6a21\u6001\u6570\u636e\u8fdb\u884c\u5206\u7c7b\u5b9e\u9a8c\u7684\u60c5\u51b5\u4e0b\u7684ROC \u66f2\u7ebf\u3002\u901a\u8fc7\u5bf9\u6bd4\u89c2\u5bdf\u516d\u4e2a\u6a21\u578b\u7684ROC \u66f2\u7ebf\u7684\u5f62\u72b6\u548c\u9762\u79ef\u6765\u8bc4\u4f30\u4e0d\u540c\u6a21\u578b\u7684\u6027\u80fd\u3002\u9760\u8fd1\u5de6\u4e0a\u89d2\u7684 ROC \u66f2\u7ebf\u8868\u793a\u5047\u9633\u6027\u7387\u63a5\u8fd10\uff0c\u771f\u9633\u6027\u7387\u63a5\u8fd1 1\uff0c\u8d8b\u8fd1\u4e8e\u5de6\u4e0a\u89d2\u7684 ROC \u66f2\u7ebf\u63d0\u793a\u6a21\u578b\u6027\u80fd\u4f18\u8d8a\u3002\u5bf9\u6bd4\u56fe 2-7 \u548c\u56fe 2-8 \u53ef\u77e5\uff0c\u4f7f\u7528\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u6570\u636e\u7684\u9aa8\u8f6c\u79fb\u75c5\u4f8b\u5206\u7c7b\u6a21\u578b\u7684\u6027\u80fd\u66f4\u4e3a\u4f18\u8d8a\u3002<\/p>\n<p>\u56fe 2-7 \u57fa\u4e8e\u5355\u6a21\u6001\u6570\u636e\u4e0d\u540c\u5206\u7c7b\u6a21\u578b\u7684\u7684ROC \u66f2\u7ebf<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-12.26.58\u202fPM.png\" alt=\"img\" \/><\/p>\n<p>\u56fe 2-8 \u57fa\u4e8e\u8de8\u6a21\u6001\u6570\u636e\u4e0d\u540c\u5206\u7c7b\u6a21\u578b\u7684\u7684ROC \u66f2\u7ebf<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-02-at-12.27.21\u202fPM.png\" alt=\"img\" \/><\/p>\n<p>\u672c\u6587\u63d0\u51fa\u7684\u8de8\u6a21\u6001\u6570\u636e\u878d\u5408\u65b9\u6cd5\u662f\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u540e\u878d\u5408\u7b56\u7565\uff0c\u76f8\u8f83\u4e8e\u524d\u878d\u5408\u7b56\u7565\u5177\u6709\u66f4\u597d\u7684\u6027\u80fd\u3002\u5b9e\u9a8c\u8868\u660e\uff0c\u65e0\u8bba\u91c7\u7528\u524d\u878d\u5408\u6216\u540e\u878d\u5408\u7b56\u7565\uff0c\u57fa\u4e8e\u8de8\u6a21\u6001\u878d\u5408\u6570\u636e\u7684\u5b9e\u9a8c\u90fd\u8868\u73b0\u51fa\u4e86\u663e\u8457\u7684\u4f18\u52bf\u3002\u76f8\u8f83\u4e8e\u591a\u6a21\u6001\u6570\u636e\uff0c\u5355\u4e00\u6a21\u6001\u6570\u636e\u6240\u6355\u83b7\u7684\u7279\u5f81\u8f83\u4e3a\u5355\u4e00\uff08\u5982\u4ec5\u6709\u7ed3\u6784\u4fe1\u606f\uff09 \uff0c\u53ef\u80fd\u7531\u4e8e\u7f3a\u4e4f\u5173\u952e\u548c\u5168\u9762\u7684\u7279\u5f81\u4fe1\u606f\u5bfc\u81f4\u6a21\u578b\u6027\u80fd\u4e0d\u4f73\uff0c\u800c\u8de8\u6a21\u6001\u878d\u5408\u65b9\u6cd5\u5219\u80fd\u4ece\u4e0d\u540c\u6a21\u6001\u4e2d\u83b7\u53d6\u66f4\u591a\u7684\u6709\u6548\u7279\u5f81\uff0c\u5e76\u5c06\u5176\u878d\u5408\uff0c\u4ece\u800c\u63d0\u9ad8\u51c6\u786e\u7387\u3002<\/p>\n<p><strong>2.4. \u8ba8\u8bba<\/strong><\/p>\n<p>\u672c\u9879\u76ee\u65e8\u5728\u8bc4\u4f30\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u6570\u636e\u878d\u5408\u65b9\u6cd5\u662f\u5426\u53ef\u4ee5\u8de8\u6a21\u6001\u878d\u5408 CT\u3001MRI \u548c PET \u56fe\u50cf\u7684\u6709\u6548\u7279\u5f81\uff0c\u4ee5\u5bf9\u4e73\u817a\u764c\u60a3\u8005\u8fdb\u884c\u662f\u5426\u53d1\u751f\u9aa8\u8f6c\u79fb\u7684\u8bc4\u4f30\u3002\u672c\u9879\u76ee\u63d0\u51fa\u4e86\u4e00\u4e2a\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u878d\u5408\u56fe\u50cf\u6570\u636e\u6846\u67b6\uff0c\u7528\u4e8e\u5bf9\u9aa8\u8f6c\u79fb\u75c5\u53d8\u8fdb\u884c\u5206\u7c7b\uff0c\u5305\u62ec\u76ee\u6807\u75c5\u53d8\u533a\u57df\u68c0\u6d4b\u3001\u7279\u5f81\u6784\u9020\u53ca\u878d\u5408\u5f62\u6210\u53ef\u89e3\u91ca\u7684\u7ed3\u6784\u5316\u6570\u636e\u4ee5\u53ca\u8de8\u6a21\u6001\u878d\u5408\u6570\u636e\u5206\u7c7b\u6b65\u9aa4\u3002\u878d\u5408 CT\u3001MRI \u548c PET \u7684\u8f6c\u6362\u7279\u5f81\u6570\u636e\u80fd\u591f\u5145\u5206\u5229\u7528\u5404\u4e2a\u6a21\u6001\u4e2d\u7684\u4fe1\u606f\uff0c\u4e3a\u5206\u7c7b\u4efb\u52a1\u63d0\u4f9b\u66f4\u591a\u7684\u6570\u636e\u652f\u6301\uff0c\u5e76\u589e\u52a0\u8f85\u52a9\u5224\u65ad\u7684\u7279\u5f81\u7ebf\u7d22\u3002\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u75c5\u4f8b\u5206\u7c7b\u5b9e\u9a8c\u8bc1\u660e\uff0c\u672c\u9879\u76ee\u63d0\u51fa\u7684\u8de8\u6a21\u6001\u878d\u5408\u6570\u636e\u663e\u8457\u63d0\u9ad8\u4e86\u5bf9\u5f71\u50cf\u5e8f\u5217\u8fdb\u884c\u4e8c\u5206\u7c7b\u4efb\u52a1\u7684\u6027\u80fd\uff0c\u76f8\u8f83\u4e8e\u5355\u6a21\u6001\u6570\u636e\uff0c\u5e73\u5747\u51c6\u786e\u7387\u548c AUC\u5206\u522b\u63d0\u9ad8\u4e867.9%\u548c 8.5%\u3002<\/p>\n<p>\u5982\u8868 2-2 \u6240\u793a\uff0c\u5355\u6a21\u6001\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u5728 PET \u56fe\u50cf\u4e2d\u7684\u68c0\u6d4b\u7cbe\u5ea6\u8f83\u9ad8\uff0c\u800c\u5728 CT\u548c MRI \u56fe\u50cf\u4e2d\u7684\u7cbe\u5ea6\u76f8\u5bf9\u8f83\u4f4e\u3002\u56fe 2-4 \u5c55\u793a\u4e86\u5728 YOLOv5 \u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u4e2d\u4e0d\u540c\u5355\u4e00\u8f93\u5165\u56fe\u50cf\u6a21\u6001\u4e0b\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u68c0\u6d4b\u6548\u679c\u7684\u53ef\u89c6\u5316\u56fe\u50cf\u3002\u5206\u6790\u5404\u79cd\u56fe\u50cf\u6a21\u6001\u4fe1\u606f\u53ef\u77e5\uff0cCT \u548c MRI \u56fe\u50cf\u4e0d\u4ec5\u5305\u542b\u75c5\u7076\u7684\u89e3\u5256\u4fe1\u606f\uff0c\u8fd8\u5305\u542b\u4e86\u4eba\u4f53\u6b63\u5e38\u7ec4\u7ec7\u7684\u7ed3\u6784\u4fe1\u606f\uff0c\u800c\u540e\u8005\u53ef\u80fd\u4f1a\u5e72\u6270\u5bab\u9888\u764c\u75c5\u53d8\u533a\u57df\u7279\u5f81\u7684\u8bc6\u522b\u548c\u68c0\u6d4b\uff0c\u5bfc\u81f4\u5355\u4e00CT\u6216MRI\u6a21\u6001\u4e0b\u51fa\u73b0\u6f0f\u68c0\u73b0\u8c61\u3002\u4e0e\u4e4b\u4e0d\u540c\uff0cPET\u56fe\u50cf\u5c55\u793a\u7684\u662f\u7ec4\u7ec7\u4ee3\u8c22\u4fe1\u606f\u3002\u9aa8\u8f6c\u79fb\u75c5\u7076\u901a\u5e38\u4f34\u968f\u7740\u9891\u7e41\u7684\u6210\u9aa8\u548c\u7834\u9aa8\u6d3b\u52a8\uff0c\u5728 PET \u5f71\u50cf\u4e2d\u5448\u9ad8\u4ee3\u8c22\uff0c\u800c\u6b63\u5e38\u9aa8\u7ec4\u7ec7\u7684\u4ee3\u8c22\u76f8\u5bf9\u8f83\u7f13\u6162\uff0c\u901a\u5e38\u4e0d\u4f1a\u663e\u793a\u5728\u56fe\u50cf\u4e2d\u3002\u56e0\u6b64\uff0cPET \u5bf9\u80cc\u666f\u7ec4\u7ec7\u7684\u5e72\u6270\u8f83 CT\u3001MRI \u66f4\u4e0d\u654f\u611f\uff0c\u6709\u52a9\u4e8e\u76ee\u6807\u68c0\u6d4b\u6a21\u578b\u8bc6\u522b\u5f02\u5e38\u4ee3\u8c22\u533a\u57df\u3002\u56e0\u65e9\u671f\u65e0\u75c7\u72b6\u9aa8\u8f6c\u79fb\u75c5\u7076\u901a\u5e38\u4f53\u79ef\u8f83\u5c0f\uff0c\u53ef\u80fd\u56e0\u76ee\u6807\u533a\u57df\u9762\u79ef\u8fc7\u5c0f\u5f71\u54cd\u76ee\u6807\u68c0\u6d4b\u7ed3\u679c\u3002\u5728\u6a21\u578b\u5904\u7406\u8fc7\u7a0b\u4e2d\uff0c\u6c60\u5316\uff08pooling\uff09\u64cd\u4f5c\u53ef\u80fd\u5bfc\u81f4\u7279\u5f81\u6216\u56fe\u50cf\u4fe1\u606f\u7684\u635f\u5931\uff0c\u4ece\u800c\u9020\u6210\u7279\u5f81\u7f3a\u5931\u3002\u4e3a\u4e86\u514b\u670d\u8fd9\u4e00\u95ee\u9898\uff0c\u540e\u7eed\u7814\u7a76\u53ef\u4ee5\u5173\u6ce8\u63d0\u9ad8\u6a21\u578b\u5728\u5904\u7406\u5c0f\u76ee\u6807\u533a\u57df\u65f6\u7684\u6027\u80fd\u3002<\/p>\n<p>\u8868 2-4 \u548c\u8868 2-6 \u5c55\u793a\u4e86\u5728\u5404\u79cd\u5206\u7c7b\u6a21\u578b\u4e2d\uff0c\u57fa\u4e8e\u8de8\u6a21\u6001\u7ed3\u6784\u5316\u6570\u636e\u5728\u4e94\u6298\u4ea4\u53c9\u9a8c\u8bc1\u4e0b\u7684\u5206\u7c7b\u6027\u80fd\u3002\u901a\u8fc7\u5bf9\u6bd4\u5206\u6790\u53d1\u73b0\uff0c\u76f8\u5bf9\u4e8e\u57fa\u4e8e C3D \u7684\u524d\u878d\u5408\u5206\u7c7b\u6a21\u578b\uff0c\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u540e\u878d\u5408\u7b56\u7565\u5728\u6027\u80fd\u65b9\u9762\u6709\u6240\u63d0\u9ad8\u3002\u533b\u5b66\u5f71\u50cf\u6570\u636e\u6709\u6570\u636e\u91cf\u8f83\u5c11\u3001\u7ef4\u5ea6\u9ad8\u3001\u7ed3\u6784\u590d\u6742\u4ee5\u53ca\u6837\u672c\u8bc6\u522b\u96be\u5ea6\u5927\u7b49\u7279\u70b9\uff0c\u8fd9\u5bfc\u81f4\u5c06\u7279\u5f81\u63d0\u53d6\u4efb\u52a1\u4ea4\u7531\u6a21\u578b\u5b8c\u6210\uff0c\u76f4\u63a5\u8f93\u5165\u539f\u59cb\u6570\u636e\u6216\u7ecf\u8fc7\u7b80\u5355\u9884\u5904\u7406\u7684\u6570\u636e\uff0c\u8ba9\u6a21\u578b\u81ea\u4e3b\u8fdb\u884c\u7279\u5f81\u63d0\u53d6\u5e76\u751f\u6210\u6700\u7ec8\u8f93\u51fa\u7684\u8fd9\u79cd\u7aef\u5230\u7aef\u7684\u524d\u878d\u5408\u65b9\u6cd5\u6548\u679c\u4e0d\u5c3d\u5982\u4eba\u610f\u3002\u7531\u4e8e\u60a3\u8005\u5728\u4f53\u578b\u3001\u8eab\u9ad8\u7b49\u65b9\u9762\u7684\u4e2a\u4f53\u5dee\u5f02\uff0c\u4e00\u4e2a\u56fe\u50cf\u5e8f\u5217\u5185\u7684CT\u548cMRI\u5207\u7247\u6570\u91cf\u4e5f\u6709\u6240\u4e0d\u540c\u3002\u56e0\u6b64\uff0c\u9700\u8981\u5bf9\u56fe\u50cf\u8fdb\u884c\u5f52\u4e00\u5316\u5904\u7406\uff0c\u5c06\u5176\u8f6c\u6362\u4e3a\u7edf\u4e00\u7684\u6807\u51c6\u683c\u5f0f\uff0c\u5982\u8c03\u6574\u5230\u76f8\u540c\u5c3a\u5bf8\u3001\u4fee\u6b63\u5207\u5272\u540e\u56fe\u50cf\u4e2d\u5fc3\u7684\u4f4d\u7f6e\u7b49\u3002\u5f52\u4e00\u5316\u64cd\u4f5c\u65e8\u5728\u5bf9\u6570\u636e\u8fdb\u884c\u7edf\u4e00\u683c\u5f0f\u5316\u548c\u538b\u7f29\uff0c\u4f46\u8fd9\u53ef\u80fd\u4f1a\u5bfc\u81f4\u56fe\u50cf\u672a\u5bf9\u9f50\u3001\u56fe\u50cf\u4e0e\u7279\u5f81\u9519\u4f4d\u3001\u6570\u636e\u538b\u7f29\u8fc7\u5ea6\u4ee5\u53ca\u7279\u5f81\u4e22\u5931\u7b49\u95ee\u9898\u3002\u56e0\u6b64\uff0c\u91c7\u7528\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u540e\u878d\u5408\u7b56\u7565\u53ef\u80fd\u66f4\u5408\u9002\u672c\u9879\u76ee\u3002\u524d\u878d\u5408\u6240\u91c7\u7528\u7684 C3D\u5206\u7c7b\u6a21\u578b\u662f\u4e00\u79cd\u5728\u4e09\u7ef4\u6570\u636e\u4e0a\u8fdb\u884c\u5206\u6790\u7684\u7f51\u7edc\u6a21\u578b\u3002\u4e09\u7ef4\u6570\u636e\u5177\u6709\u5c3a\u5ea6\u9ad8\u3001\u7ef4\u5ea6\u5927\u4ee5\u53ca\u4fe1\u606f\u7a00\u758f\u7b49\u7279\u70b9\u3002\u5c3d\u7ba1 C3D \u7f51\u7edc\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u589e\u52a0\u4e86\u4e00\u4e2a\u7ef4\u5ea6\u7684\u4fe1\u606f\uff0c\u4f46\u540c\u65f6\u4e5f\u63d0\u9ad8\u4e86\u7b97\u6cd5\u5206\u6790\u7684\u590d\u6742\u6027\uff0c\u7279\u522b\u662f\u5728\u6a21\u578b\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\uff0c\u5360\u7528\u4e86\u5927\u91cf\u663e\u5b58\u7b49\u786c\u4ef6\u8d44\u6e90\uff0c\u53ef\u80fd\u5bfc\u81f4\u6279\u5f52\u4e00\u5316\u4e0d\u7406\u60f3\u548c\u7f51\u7edc\u6536\u655b\u4e0d\u5b8c\u5168\u7684\u95ee\u9898\u3002\u4e0e C3D \u76f8\u6bd4\uff0c\u672c\u6587\u63d0\u51fa\u7684\u4e8c\u9636\u6bb5\u540e\u878d\u5408\u65b9\u6cd5\u5b9e\u73b0\u4e86\u7279\u5f81\u538b\u7f29\uff0c\u63d0\u53d6\u7f6e\u4fe1\u5ea6\u8fd9\u79cd\u53ef\u89e3\u91ca\u7684\u7279\u5f81\uff0c\u5e76\u53bb\u9664\u4e86\u65e0\u5173\u7684\u7a00\u758f\u7279\u5f81\u3002\u5728\u6709\u9650\u7684\u786c\u4ef6\u8d44\u6e90\u548c\u6570\u636e\u91cf\u7684\u9650\u5236\u4e0b\uff0c\u8fd9\u79cd\u65b9\u6cd5\u80fd\u66f4\u597d\u5730\u5b66\u4e60\u6570\u636e\u7279\u5f81\uff0c\u8d77\u5230\u4e86\u7c7b\u4f3c\u6b63\u5219\u5316\uff08\u901a\u8fc7\u5728\u635f\u5931\u51fd\u6570\u4e2d\u6dfb\u52a0\u7ea6\u675f\uff0c\u89c4\u8303\u6a21\u578b\u5728\u540e\u7eed\u8fed\u4ee3\u8fc7\u7a0b\u4e2d\u907f\u514d\u8fc7\u5ea6\u62df\u5408\uff09\u7684\u6548\u679c\u3002<\/p>\n<p>\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u75c5\u7076\u5728\u4ee3\u8c22\u548c\u7ed3\u6784\u65b9\u9762\u90fd\u8f83\u6b63\u5e38\u9aa8\u7ec4\u7ec7\u663e\u8457\u4e0d\u540c\u3002\u56e0\u6b64\uff0c\u901a\u8fc7\u878d\u5408 CT\u3001MRI \u548c PET \u56fe\u50cf\u7684\u7279\u5f81\u4fe1\u606f\uff0c\u5b9e\u73b0\u89e3\u5256\u548c\u529f\u80fd\u4fe1\u606f\u7684\u8de8\u6a21\u6001\u878d\u5408\uff0c\u80fd\u66f4\u6709\u6548\u5730\u5b8c\u6210\u5206\u7c7b\u4efb\u52a1\uff0c\u5e2e\u52a9\u8bca\u65ad\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u3002\u7136\u800c\uff0c\u7efc\u5408\u5206\u6790\u5168\u8eab CT\u3001MRI \u548cPET\u56fe\u50cf\u4fe1\u606f\u9700\u8981\u533b\u5e08\u6295\u5165\u5927\u91cf\u65f6\u95f4\uff0c\u4e14\u5b58\u5728\u8f83\u5927\u7684\u89c2\u5bdf\u8005\u95f4\u5dee\u5f02\u3002\u4e00\u65e6\u53d1\u751f\u6f0f\u8bca\uff0c\u4f1a\u5bfc\u81f4\u4e25\u91cd\u540e\u679c\u3002\u5229\u7528\u8ba1\u7b97\u673a\u8f85\u52a9\u533b\u5e08\u5224\u65ad\u4e73\u817a\u764c\u662f\u5426\u53d1\u751f\u9aa8\u8f6c\u79fb\u4e0d\u4ec5\u53ef\u4ee5\u8282\u7701\u6210\u672c\u548c\u65f6\u95f4\uff0c\u8fd8\u80fd\u63d0\u4f9b\u66f4\u52a0\u5ba2\u89c2\u7684\u8bc4\u4f30\u6807\u51c6\u3002\u8ba1\u7b97\u673a\u8f85\u52a9\u8bca\u65ad\u5de5\u5177\u53ef\u4ee5\u7efc\u5408\u591a\u6a21\u6001\u56fe\u50cf\u7684\u7ed3\u679c\u8fdb\u884c\u7279\u5f81\u8f6c\u6362\u548c\u5206\u6790\uff0c\u9884\u9632\u6f0f\u8bca\u7684\u53d1\u751f\u3002\u56e0\u6b64\uff0c\u5728\u672a\u6765\u7684\u7814\u7a76\u4e2d\uff0c\u53ef\u4ee5\u91cd\u70b9\u5173\u6ce8\u5f00\u53d1\u6b64\u7c7b\u8ba1\u7b97\u673a\u8f85\u52a9\u8bca\u65ad\u7cfb\u7edf\uff0c\u4ee5\u63d0\u9ad8\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u8bca\u65ad\u7684\u51c6\u786e\u6027\u548c\u6548\u7387\u3002<\/p>\n<p>\u8fd9\u9879\u7814\u7a76\u4ecd\u5b58\u5728\u4e00\u4e9b\u5c40\u9650\u6027\u3002\u4ece\u5355\u4e2a\u5f71\u50cf\u6a21\u6001\u4e2d\u63d0\u53d6\u7684\u7279\u5f81\u8f83\u4e3a\u5355\u4e00\uff0c\u4ec5\u6709\u7f6e\u4fe1\u533a\u95f4\uff0c\u53ef\u4ee5\u5728\u540e\u7eed\u7684\u8bad\u7ec3\u4e2d\u53ef\u4ee5\u4ece\u4e34\u5e8a\u89d2\u5ea6\u51fa\u53d1\u52a0\u5165\u66f4\u591a\u53ef\u80fd\u5f71\u54cd\u9aa8\u8f6c\u79fb\u5224\u65ad\u7684\u56e0\u7d20\u4f5c\u4e3a\u5206\u7c7b\u7279\u5f81\uff0c\u5982\u68c0\u6d4b\u76ee\u6807\u7684\u9762\u79ef\uff0c\u6216\u589e\u52a0\u4e2d\u8f74\u9aa8\u68c0\u51fa\u76ee\u6807\u7684\u6743\u91cd\u3002\u56e0\u672c\u9879\u7814\u7a76\u5177\u6709\u591a\u6a21\u6001\u5f71\u50cf\u6570\u636e\u7684\u75c5\u4f8b\u91cf\u4e0d\u591f\uff0c\u672a\u6765\u53ef\u4ee5\u5c1d\u8bd5\u9664\u4e94\u6298\u4ea4\u53c9\u9a8c\u8bc1\u4e4b\u5916\u5176\u4ed6\u7684\u6a21\u578b\u8bad\u7ec3\u65b9\u6cd5\u4ee5\u964d\u4f4e\u6570\u636e\u91cf\u5bf9\u5206\u7c7b\u6a21\u578b\u6027\u80fd\u7684\u5f71\u54cd\u3002<\/p>\n<p><strong>2.5. \u7ed3\u8bba<\/strong><\/p>\n<p>\u672c\u9879\u76ee\u63d0\u51fa\u4e86\u4e00\u79cd\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u6570\u636e\u878d\u5408\u65b9\u6cd5\u8fdb\u884c\u5206\u7c7b\u4efb\u52a1\u7684\u6df1\u5ea6\u5b66\u4e60\u6846\u67b6\uff0c\u7528\u4e8e\u5224\u65ad\u662f\u5426\u53d1\u751f\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u3002\u9996\u5148\u72ec\u7acb\u5bf9\u4e0d\u540c\u6a21\u6001\u7684\u533b\u5b66\u56fe\u50cf\u6570\u636e\u8fdb\u884c\u80bf\u7624\u68c0\u6d4b\uff0c\u6839\u636e\u76ee\u6807\u68c0\u6d4b\u7ed3\u679c\u8fdb\u884c\u7279\u5f81\u6784\u9020\uff0c\u5e76\u5c06\u5176\u7ec4\u7ec7\u6210\u7ed3\u6784\u5316\u6570\u636e\u7684\u5f62\u5f0f\uff0c\u5b8c\u6210\u4ece\u975e\u7ed3\u6784\u5316\u6570\u636e\u7279\u5f81\u5230\u7ed3\u6784\u5316\u6570\u636e\u7279\u5f81\u7684\u8f6c\u6362\u4e0e\u878d\u5408\u3002\u6700\u7ec8\uff0c\u5c06\u7ed3\u6784\u5316\u6570\u636e\u7279\u5f81\u8f93\u5165\u5206\u7c7b\u5668\uff0c\u8fdb\u884c\u9aa8\u8f6c\u79fb\u7684\u5206\u7c7b\u4efb\u52a1\uff0c\u5e76\u5bf9\u7167 C3D \u524d\u878d\u5408\u6a21\u578b\uff0c\u8ba8\u8bba\u4e86\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u65b9\u6cd5\u8fdb\u884c\u8de8\u6a21\u6001\u6570\u636e\u540e\u878d\u5408\u7684\u4f18\u52bf\u3002\u5927\u91cf\u7684\u5b9e\u9a8c\u8bc1\u660e\uff0c\u4f7f\u7528\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u878d\u5408\u6570\u636e\u8fdb\u884c\u5206\u7c7b\u4efb\u52a1\u7684\u6027\u80fd\u4f18\u4e8e\u57fa\u4e8e\u5355\u6a21\u6001\u6570\u636e\u7684\u5206\u7c7b\u6027\u80fd\uff1b\u4f7f\u7528\u672c\u9879\u76ee\u63d0\u51fa\u7684\u540e\u878d\u5408\u7b56\u7565\u6267\u884c\u5206\u7c7b\u4efb\u52a1\u8f83\u4f7f\u7528\u524d\u878d\u5408\u7b56\u7565\u7684\u5206\u7c7b\u6a21\u578b\uff08C3D\uff09\u7684\u6027\u80fd\u66f4\u597d\u3002<\/p>\n<p>\u672c\u9879\u76ee\u6240\u63d0\u51fa\u7684\u6280\u672f\u53ef\u7efc\u5408 CT\u3001MRI \u548c PET \u6a21\u6001\u6570\u636e\u7684\u7279\u5f81\uff0c\u5bf9\u4e73\u817a\u764c\u60a3\u8005\u662f\u5426\u53d1\u751f\u9aa8\u8f6c\u79fb\u8fdb\u884c\u5224\u65ad\uff0c\u8f85\u52a9\u4e34\u5e8a\u533b\u5e08\u8fdb\u884c\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u75c5\u7076\u7684\u7b5b\u67e5\uff0c\u5177\u5907\u5b9e\u9645\u5e94\u7528\u4ef7\u503c\uff0c\u4e5f\u4e3a\u5728\u533b\u5b66\u56fe\u50cf\u5206\u6790\u4efb\u52a1\u4e2d\u66f4\u6709\u6548\u5730\u5e94\u7528\u8de8\u6a21\u6001\u878d\u5408\u65b9\u6cd5\uff0c\u63d0\u4f9b\u4e86\u5173\u952e\u7684\u7406\u8bba\u652f\u6301\u3002<\/p>\n<p><strong>\u5168\u6587\u5c0f\u7ed3<\/strong><\/p>\n<p>\u76ee\u524d\uff0c\u533b\u5b66\u5f71\u50cf\u5b66\u7684\u89e3\u8bfb\u5927\u91cf\u4f9d\u8d56\u4e34\u5e8a\u533b\u751f\u4e2a\u4eba\u7684\u4e3b\u89c2\u8bca\u65ad\u7ecf\u9a8c\uff0c\u4eba\u5de5\u9605\u7247\u6613\u6f0f\u8bca\u5c0f\u76ee\u6807\uff0c\u96be\u4ee5\u63a8\u5e7f\u53ca\u8868\u8ff0\uff0c\u5177\u6709\u4e00\u5b9a\u7684\u5c40\u9650\u6027\u3002\u4e0e\u6b64\u76f8\u6bd4\uff0c\u4eba\u5de5\u667a\u80fd\u6280\u672f\u53ef\u4ee5\u901a\u8fc7\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\u5bf9\u5927\u91cf\u79ef\u7d2f\u7684\u5f71\u50cf\u6570\u636e\u548c\u8bca\u65ad\u6570\u636e\u8fdb\u884c\u5206\u6790\uff0c\u5b66\u4e60\u5e76\u63d0\u53d6\u6570\u636e\u4e2d\u5bf9\u75c5\u7406\u8bca\u65ad\u6709\u7528\u7684\u7279\u5f81\uff0c\u4ece\u800c\u5728\u6570\u636e\u652f\u6301\u4e0b\u505a\u51fa\u66f4\u5ba2\u89c2\u7684\u5224\u65ad\u3002\u6309\u6210\u50cf\u65b9\u5f0f\u4e0d\u540c\uff0c\u533b\u5b66\u5f71\u50cf\u6570\u636e\u53ef\u5206\u4e3a\u591a\u79cd\u6a21\u6001\uff0c\u5982B\u8d85\u3001CT\u3001MRI\u3001PET\u3002\u4e3a\u4e86\u6700\u5927\u9650\u5ea6\u6a21\u62df\u4e34\u5e8a\u533b\u751f\u7ed3\u5408\u4e0d\u540c\u6a21\u6001\u5f71\u50cf\u68c0\u67e5\u7ed3\u679c\u5f62\u6210\u8bca\u65ad\u7684\u8fc7\u7a0b\uff0c\u8bbe\u8ba1\u4eba\u5de5\u667a\u80fd\u6a21\u578b\u65f6\uff0c\u5e94\u5c06\u5404\u79cd\u5f71\u50cf\u5b66\u6a21\u6001\u7684\u7279\u5f81\u8fdb\u884c\u6709\u6548\u7684\u878d\u5408\uff0c\u5373\u672c\u9879\u76ee\u4e2d\u5e94\u7528\u7684\u8de8\u6a21\u6001\u6df1\u5ea6\u5b66\u4e60\u65b9\u6cd5\uff0c\u5145\u5206\u5229\u7528\u4e0d\u540c\u6a21\u6001\u56fe\u50cf\u7684\u72ec\u7279\u4f18\u52bf\u8bad\u7ec3\u6df1\u5ea6\u795e\u7ecf\u7f51\u7edc\uff0c\u4ece\u800c\u63d0\u9ad8\u6a21\u578b\u6027\u80fd\u3002\u672c\u9879\u76ee\u4ee5\u5bab\u9888\u764c\u548c\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u4e3a\u4f8b\uff0c\u9a8c\u8bc1\u4e86\u8de8\u6a21\u6001\u6df1\u5ea6\u5b66\u4e60\u65b9\u6cd5\u5728\u75c5\u53d8\u533a\u57df\u5b9a<br \/>\n\u4f4d\u548c\u8f85\u52a9\u8bca\u65ad\u65b9\u9762\u7684\u6027\u80fd\u3002<\/p>\n<p>\u5728\u7b2c\u4e00\u90e8\u5206\u4e2d\uff0c\u6211\u4eec\u56de\u987e\u6027\u7eb3\u5165\u4e86220\u4f8b\u6709FDG-PET\/CT\u6570\u636e\u7684\u5bab\u9888\u764c\u60a3\u8005\uff0c\u5171\u8ba1 72,602 \u5f20\u5207\u7247\u56fe\u50cf\u3002\u901a\u8fc7\u56fe\u50cf\u589e\u5f3a\u3001\u8fb9\u7f18\u68c0\u6d4b\uff0c\u5b9e\u73b0 PET \u548c CT \u56fe\u50cf\u7684 ROI\u81ea\u9002\u5e94\u5b9a\u4f4d\uff0c\u518d\u901a\u8fc7\u7f29\u653e\u3001\u96f6\u503c\u586b\u5145\u548c\u526a\u5207\u7684\u65b9\u5f0f\uff0c\u5c06\u4e24\u79cd\u6a21\u6001\u56fe\u50cf\u7684 ROI \u5bf9\u9f50\u3002\u7ecf\u8fc7\u52a0\u6743\u548c\u56fe\u50cf\u53e0\u52a0\uff0c\u8fdb\u884c\u56fe\u50cf\u878d\u5408\uff0c\u5c06\u878d\u5408\u540e\u7684\u56fe\u50cf\u4f5c\u4e3a\u76ee\u6807\u68c0\u6d4b\u7f51\u7edc\u7684\u8f93\u5165\u5c42\uff0c\u8fdb\u884c\u5bab\u9888\u764c\u75c5\u53d8\u533a\u57df\u68c0\u6d4b\u3002\u5b9e\u9a8c\u8bc1\u660e\uff0c\u76f8\u6bd4\u4f7f\u7528\u5355\u4e00 CT \u56fe\u50cf\u3001\u5355\u4e00 PET \u56fe\u50cf\u3001PET-CT \u7b80\u5355\u878d\u5408\u56fe\u50cf\u3001PET-CT \u901a\u9053\u62fc\u63a5\u878d\u5408\u56fe\u50cf\u4f5c\u4e3a\u7f51\u7edc\u8f93\u5165\uff0cPET-CT \u81ea\u9002\u5e94\u533a\u57df\u7279\u5f81\u878d\u5408\u56fe\u50cf\u663e\u8457\u63d0\u9ad8\u4e86\u5bab\u9888\u764c\u75c5\u53d8\u533a\u57df\u68c0\u6d4b\u7684\u51c6\u786e\u6027\uff0c\u76ee\u6807\u68c0\u6d4b\u7684\u5e73\u5747\u7cbe\u786e\u5ea6\uff08AP50\uff09\u5206\u522b\u63d0\u9ad8\u4e86 6.06%\u548c 8.9%\uff0c\u4e14\u6d88\u9664\u4e86\u4e00\u4e9b\u5047\u9633\u6027\u7ed3\u679c\uff0c\u5c55\u73b0\u51fa\u53ef\u89c2\u7684\u4e34\u5e8a\u5e94\u7528\u4ef7\u503c\u3002<\/p>\n<p>\u5728\u7b2c\u4e8c\u90e8\u5206\u4e2d\uff0c\u6211\u4eec\u56de\u987e\u6027\u7eb3\u5165\u4e86 233 \u4f8b\u4e73\u817a\u764c\u60a3\u8005\uff0c\u6bcf\u4f8b\u6837\u672c\u6570\u636e\u5305\u542b CT\u3001MRI\u3001\u6216 PET \u4e00\u81f3\u4e09\u79cd\u6a21\u6001\u7684\u5168\u8eab\u5f71\u50cf\u6570\u636e\uff0c\u5171\u6709 3051 \u5f20 CT \u5207\u7247\uff0c3543 \u5f20 MRI\u5207\u7247\uff0c1818 \u5f20 PET \u5207\u7247\u3002\u9996\u5148\u8bad\u7ec3 YOLOv5 \u76ee\u6807\u68c0\u6d4b\u7f51\u7edc\uff0c\u5bf9\u6bcf\u79cd\u5355\u4e00\u6a21\u6001\u56fe\u50cf\u4e2d\u7684\u9aa8\u8f6c\u79fb\u75c5\u7076\u8fdb\u884c\u76ee\u6807\u68c0\u6d4b\u3002\u7edf\u8ba1\u6bcf\u4e2a\u5f71\u50cf\u5e8f\u5217\u4e2d\u542b\u6709\u68c0\u51fa\u9aa8\u8f6c\u79fb\u75c5\u7076\u7684\u4e2a\u6570\u548c\u7f6e\u4fe1\u5ea6\uff0c\u5c06\u6bcf\u4e2a\u7f6e\u4fe1\u533a\u95f4\u5185\u542b\u6709\u76ee\u6807\u68c0\u6d4b\u6846\u7684\u767e\u5206\u6bd4\u4f5c\u4e3a\u7ed3\u6784\u5316\u533b\u7597\u7279\u5f81\u6570\u636e\u3002\u91c7\u7528\u7ea7\u8054\u65b9\u5f0f\u878d\u5408\u4e09\u79cd\u6a21\u6001\u7684\u7ed3\u6784\u5316\u7279\u5f81\uff0c\u5f97\u5230\u5177\u6709\u53ef\u89e3\u91ca\u6027\u7684\u7ed3\u6784\u5316\u533b\u7597\u6570\u636e\uff0c\u518d\u7528\u5206\u7c7b\u6a21\u578b\u8fdb\u884c\u5206\u7c7b\uff0c\u9884\u6d4b\u662f\u5426\u53d1\u751f\u9aa8\u8f6c\u79fb\u3002\u5b9e\u9a8c\u8bc1\u660e\uff0c\u76f8\u8f83\u4e8e\u5355\u6a21\u6001\u6570\u636e\uff0c\u8de8\u6a21\u6001\u878d\u5408\u6570\u636e\u663e\u8457\u63d0\u9ad8\u4e86\u4e73\u817a\u764c\u9aa8\u8f6c\u79fb\u8bca\u65ad\u4efb\u52a1\u7684\u6027\u80fd\uff0c\u5e73\u5747\u51c6\u786e\u7387\u548c AUC \u5206\u522b\u63d0\u9ad8\u4e86 7.9%\u548c 8.5%\uff0c\u89c2\u5bdf ROC \u66f2\u7ebf\u548c PR \u66f2\u7ebf\u7684\u5f62\u72b6\u548c\u9762\u79ef\u4e5f\u6709\u76f8\u540c\u7684\u5b9e\u9a8c\u7ed3\u8bba\u3002\u5728\u4e0d\u540c\u7684\u5206\u7c7b\u6a21\u578b\uff08SVM\u3001AdaBoost\u3001RandomForest\u3001LightGBM\u3001GBDT\uff09\u4e2d\uff0c\u4f7f\u7528\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u8de8\u6a21\u6001\u6570\u636e\uff0c\u76f8\u6bd4\u5355\u6a21\u6001\u6570\u636e\uff0c\u5bf9\u4e8e\u9aa8\u8f6c\u79fb\u75c5\u4f8b\u7684\u5206\u7c7b\u6027\u80fd\u66f4\u4e3a\u4f18\u8d8a\u3002\u800c\u76f8\u8f83\u4e8e\u57fa\u4e8e C3D \u7684\u524d\u878d\u5408\u5206\u7c7b\u6a21\u578b\uff0c\u57fa\u4e8e\u7279\u5f81\u8f6c\u6362\u7684\u540e\u878d\u5408\u7b56\u7565\u5728\u5206\u7c7b\u4efb\u52a1\u65b9\u9762\u7684\u6027\u80fd\u66f4\u4f18\u3002<\/p>\n<p>\u7efc\u4e0a\u6240\u8ff0\uff0c\u672c\u6587\u57fa\u4e8e\u4eba\u5de5\u667a\u80fd\u6df1\u5ea6\u5b66\u4e60\u7b97\u6cd5\uff0c\u9488\u5bf9\u4e0d\u540c\u6a21\u6001\u533b\u5b66\u56fe\u50cf\u7684\u7279\u5f81\u5dee\u5f02\u4e0e\u4e92\u8865\u6027\uff0c\u8fdb\u884c\u591a\u6a21\u6001\u533b\u5b66\u5f71\u50cf\u6570\u636e\u7684\u8de8\u6a21\u6001\u878d\u5408\uff0c\u63d0\u9ad8\u4e86\u6a21\u578b\u7684\u80bf\u7624\u68c0\u6d4b\u548c\u5206\u7c7b\u6027\u80fd\uff0c\u68c0\u6d4b\u6a21\u578b\u548c\u5206\u7c7b\u6a21\u578b\u53ef\u4ee5\u8f85\u52a9\u5f71\u50cf\u5b66\u9605\u7247\u8fc7\u7a0b\uff0c\u5177\u6709\u663e\u8457\u7684\u4e34\u5e8a\u5b9e\u9645\u5e94\u7528\u4ef7\u503c\u3002<\/p>\n<p><strong>\u53c2\u8003\u6587\u732e<\/strong><\/p>\n[1] \u9648\u601d\u6e90, \u8c2d\u827e\u8fea, \u9b4f\u53cc\u5251, \u76d6\u73c2\u73c2. \u57fa\u4e8e\u533a\u5757\u94fe\u7684\u533b\u7597\u5f71\u50cf\u6570\u636e\u4eba\u5de5\u667a\u80fd\u68c0\u6d4b\u6a21\u578b[J] . \u7f51\u7edc\u5b89\u5168\u4e0e\u6570\u636e\u6cbb\u7406, 2022, 41(10): 21-25.<br \/>\n[2] Dong X, Wu D. A rare cause of peri-esophageal cystic lesion[J . Gastroenterology, 2023, 164(2): 191-193.<br \/>\n[3] Arbyn M, Weiderpass E, Bruni L, et al. Estimates of incidence and mortality of cervical cancer in 2018: a worldwide analysis[J . The Lancet Global Health, 2020, 8(2): e191-e203.<br \/>\n[4] Marth C, Landoni F, Mahner S, et al. Cervical cancer: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up[J . Annals of Oncology, 2017, 28:iv72-iv83.<br \/>\n[5] Gold M A. PET in Cervical Cancer\u2014Implications forStaging,&#8217;Treatment Planning, Assessment of Prognosis, and Prediction of Response[J . Journal of the National Comprehensive Cancer Network, 2008, 6(1): 37-45.<br \/>\n[6] Gandy N, Arshad M A, Park W H E, et al. FDG-PET imaging in cervical cancer[C] .Seminars in nuclear medicine. WB Saunders, 2019, 49(6): 461-470.<br \/>\n[7] Grigsby P W. PET\/CT imaging to guide cervical cancer therapy[J . Future Oncology, 2009, 5(7): 953-958.<br \/>\n[8] Mirpour S, Mhlanga J C, Logeswaran P, et al. The role of PET\/CT in the management of cervical cancer[J . American Journal of Roentgenology, 2013, 201(2): W192-W205.<br \/>\n[9] LeCun Y, Bengio Y, Hinton G. Deep learning[J . nature, 2015, 521(7553): 436-444.<br \/>\n[10] Szeliski R. Computer vision: algorithms and applications Springer Science &amp; Business Media[J] . 2010.<br \/>\n[11] Ma B, Yin X, Wu D, et al. End-to-end learning for simultaneously generating decision<br \/>\nmap and multi-focus image fusion result[J . Neurocomputing, 2022, 470: 204-216.<br \/>\n[12] Anwar S M, Majid M, Qayyum A, et al. Medical image analysis using convolutional<br \/>\nneural networks: a review[J . Journal of medical systems, 2018, 42: 1-13.<br \/>\n[13] Ma B, Ban X, Huang H, et al. Deep learning-based image segmentation for al-la alloy<br \/>\nmicroscopic images[J . Symmetry, 2018, 10(4): 107.<br \/>\n[14] Li Z, He J, Zhang X, et al. Toward high accuracy and visualization: An interpretable feature extraction method based on genetic programming and non-overlap degree[C] . 2020 IEEE International Conference on Bioinformatics and Biomedicine (BIBM). IEEE, 2020: 299-304.<br \/>\n[15] Rumelhart D E, Hinton G E, Williams R J. Learning representations by back-propagating errors[J] . nature, 1986, 323(6088): 533-536.<br \/>\n[16] He K , Gkioxari G , Dollar P , et al. Mask R-CNN[C . International Conference on Computer Vision. IEEE Computer Society, 2017, pp. 2980-2988.<br \/>\n[17] Ma B, Wei X, Liu C, et al. Data augmentation in microscopic images for material data mining[J] . npj Computational Materials, 2020, 6(1): 125.<br \/>\n[18] Ma B, Zhu Y, Yin X, et al. Sesf-fuse: An unsupervised deep model for multi-focus image fusion[J] . Neural Computing and Applications, 2021, 33: 5793-5804.<br \/>\n[19] Kermany D S, Goldbaum M, Cai W, et al. Identifying medical diagnoses and treatable diseases by image-based deep learning[J . cell, 2018, 172(5): 1122-1131. e9.<br \/>\n[20] Hyun S H, Ahn M S, Koh Y W, et al. A machine-learning approach using PET-based radiomics to predict the histological subtypes of lung cancer[J . Clinical nuclear medicine, 2019, 44(12): 956-960.<br \/>\n[21] Chilamkurthy S, Ghosh R, Tanamala S, et al. Deep learning algorithms for detection of critical findings in head CT scans: a retrospective study[J . The Lancet, 2018, 392(10162): 2388-2396.<br \/>\n[22] Chen C, Xiao R, Zhang T, et al. Pathological lung segmentation in chest CT images based on improved random walker[J] . Computer methods and programs in biomedicine, 2021, 200: 105864.<br \/>\n[23] Chen C, Zhou K, Zha M, et al. An effective deep neural network for lung lesions segmentation from COVID-19 CT images[J] . IEEE Transactions on Industrial Informatics, 2021, 17(9): 6528-6538.<br \/>\n[24] Hill D L G, Batchelor P G, Holden M, et al. Medical image registration[J] . Physics in medicine &amp; biology, 2001, 46(3): R1.<br \/>\n[25] Du J, Li W, Lu K, et al. An overview of multi-modal medical image fusion[J] .Neurocomputing, 2016, 215: 3-20.<br \/>\n[26] Watanabe H, Ariji Y, Fukuda M, et al. Deep learning object detection of maxillary cyst-like lesions on panoramic radiographs: preliminary study[J] . Oral radiology, 2021, 37: 487-493.<br \/>\n[27] Mattes D, Haynor D R, Vesselle H, et al. PET-CT image registration in the chest using free-form deformations[J] . IEEE transactions on medical imaging, 2003, 22(1): 120-128.<br \/>\n[28] Maqsood S, Javed U. Multi-modal medical image fusion based on two-scale image decomposition and sparse representation[J] . Biomedical Signal Processing and Control, 2020, 57: 101810.<br \/>\n[29] Elakkiya R, Subramaniyaswamy V, Vijayakumar V, et al. Cervical cancer diagnostics healthcare system using hybrid object detection adversarial networks[J] . IEEE Journal of Biomedical and Health Informatics, 2021, 26(4): 1464-1471.<br \/>\n[30] Al-Ameen Z, Sulong G, Gapar M D, et al. Reducing the Gaussian blur artifact from CT medical images by employing a combination of sharpening filters and iterative deblurring algorithms[J] . Journal of Theoretical and Applied Information Technology, 2012, 46(1): 31-36.<br \/>\n[31] Canny J. A computational approach to edge detection[J] . IEEE Transactions on pattern analysis and machine intelligence, 1986 (6): 679-698.<br \/>\n[32] Jocher G, Stoken A, Borovec J, et al. ultralytics\/yolov5: v5. 0-YOLOv5-P6 1280 models, AWS, Supervise. ly and YouTube integrations[J . Zenodo, 2021.<br \/>\n[33] Wang C Y, Liao H Y M, Wu Y H, et al. CSPNet: A new backbone that can enhance learning capability of CNN[C . Proceedings of the IEEE\/CVF conference on computer vision and pattern recognition workshops, 2020: 390-391.<br \/>\n[34] Liu S, Qi L, Qin H, et al. Path aggregation network for instance segmentation[C .<br \/>\nProceedings of the IEEE conference on computer vision and pattern recognition, 2018:<br \/>\n8759-8768.<br \/>\n[35] Bochkovskiy A, Wang C Y, Liao H Y M. Yolov4: Optimal speed and accuracy of object detection[J . arXiv preprint arXiv, 2020, 2004: 10934.<br \/>\n[36] He K, Zhang X, Ren S, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J . IEEE transactions on pattern analysis and machine intelligence, 2015, 37(9): 1904-1916.<br \/>\n[37] Lee S I, Atri M. 2018 FIGO staging system for uterine cervical cancer: enter cross-sectional imaging[J . Radiology, 2019, 292(1): 15-24.<br \/>\n[38] Bochkovskiy A, Wang C Y, Liao H Y M. Yolov4: Optimal speed and accuracy of object detection[J . arXiv preprint arXiv:2004.10934, 2020.<br \/>\n[39] Smith A R. Color gamut transform pairs[J . ACM Siggraph Computer Graphics, 1978, 12(3): 12-19.<br \/>\n[40] Ioffe S, Szegedy C. Batch normalization: Accelerating deep network training by<br \/>\nreducing internal covariate shift[C . International conference on machine learning.<br \/>\npmlr, 2015: 448-456.<br \/>\n[41] Elfwing S, Uchibe E, Doya K. Sigmoid-weighted linear units for neural network function approximation in reinforcement learning[J . Neural Networks, 2018, 107: 3-11.<br \/>\n[42] Lin T Y, Goyal P, Girshick R, et al. Focal loss for dense object detection[C] .Proceedings of the IEEE international conference on computer vision. 2017: 2980-2988.<br \/>\n[43] Zhang S, Chi C, Yao Y, et al. Bridging the gap between anchor-based and anchor-free detection via adaptive training sample selection[C . Proceedings of the IEEE\/CVF conference on computer vision and pattern recognition. 2020: 9759-9768.<br \/>\n[44] Ren S, He K, Girshick R, et al. Faster r-cnn: Towards real-time object detection with region proposal networks[J . Advances in neural information processing systems, 2015, 28.<br \/>\n[45] Cai Z, Vasconcelos N. Cascade r-cnn: Delving into high quality object detection[C] .Proceedings of the IEEE conference on computer vision and pattern recognition. 2018:6154-6162.<br \/>\n[46] Pang J, Chen K, Shi J, et al. Libra r-cnn: Towards balanced learning for object detection[C] . Proceedings of the IEEE\/CVF conference on computer vision and pattern recognition. 2019: 821-830.<br \/>\n[47] Cohen P A, Jhingran A, Oaknin A, et al. Cervical cancer[J . The Lancet, 2019, 393(10167): 169-182.<br \/>\n[48] Lee S I, Atri M. 2018 FIGO staging system for uterine cervical cancer: enter cross-sectional imaging[J] . Radiology, 2019, 292(1): 15-24.<br \/>\n[49] Coleman R E. Metastatic bone disease: clinical features, pathophysiology and treatment strategies[J . Cancer treatment reviews, 2001, 27(3): 165-176.<br \/>\n[50] Cecchini M G, Wetterwald A, Van Der Pluijm G, et al. Molecular and biological mechanisms of bone metastasis[J . EAU Update Series, 2005, 3(4): 214-226.<br \/>\n[51] Cuccurullo V, Lucio Cascini G, Tamburrini O, et al. Bone metastases radiopharmaceuticals:<br \/>\nan overview[J . Current radiopharmaceuticals, 2013, 6(1): 41-47.<br \/>\n[52] Emens L A, Davidson N E. The follow-up of breast cancer[C . Seminars in oncology. WB Saunders, 2003, 30(3): 338-348.<br \/>\n[53] Chen W Z, Shen J F, Zhou Y, et al. Clinical characteristics and risk factors for developing bone metastases in patients with breast cancer[J . Scientific reports, 2017, 7(1): 1-7.<br \/>\n[54] Batson O V. The function of the vertebral veins and their role in the spread of metastases[J . Annals of surgery, 1940, 112(1): 138.<br \/>\n[55] O\u2019Sullivan G J, Carty F L, Cronin C G. Imaging of bone metastasis: an update[J] .World journal of radiology, 2015, 7(8): 202.<br \/>\n[56] Noguchi S, Nishio M, Sakamoto R, et al. Deep learning\u2013based algorithm improved radiologists\u2019 performance in bone metastases detection on CT[J . European Radiology, 2022, 32(11): 7976-7987.<br \/>\n[57] Fan X , Zhang X , Zhang Z , et al. Deep Learning on MRI Images for Diagnosis of Lung Cancer Spinal Bone Metastasis[J . Contrast Media &amp; Molecular Imaging, 2021, 2021(1):1-9.<br \/>\n[58] Moreau N, Rousseau C, Fourcade C, et al. Deep learning approaches for bone and bone lesion segmentation on 18FDG PET\/CT imaging in the context of metastatic breast cancer[C] . 42nd Annual International Conference of the IEEE Engineering in Medicine &amp; Biology Society (EMBC). IEEE, 2020: 1532-1535.<br \/>\n[59] Boser B E, Guyon I M, Vapnik V N. A training algorithm for optimal margin classifiers[C] . Proceedings of the fifth annual workshop on Computational learning theory, 1992: 144-152.<br \/>\n[60] Freund Y, Schapire R E. A decision-theoretic generalization of on-line learning and an application to boosting[J] . Journal of computer and system sciences, 1997, 55(1): 119-139.<br \/>\n[61] Breiman L. Random forests[J . Machine learning, 2001, 45: 5-32.<br \/>\n[62] Ke G, Meng Q, Finley T, et al. Lightgbm: A highly efficient gradient boosting decision tree[J] . Advances in neural information processing systems, 2017, 30: 52.<br \/>\n[63] Chen T, Guestrin C. Xgboost: A scalable tree boosting system[C . Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining, 2016: 785-794.<br \/>\n[64] Tran D, Bourdev L, Fergus R, et al. Learning spatiotemporal features with 3d convolutional networks[C . Proceedings of the IEEE international conference on computer vision, 2015: 4489-4497.<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-03-at-9.37.38\u202fAM.png\" alt=\"img\" \/><\/p>\n<p><strong>\u6587\u732e\u7efc\u8ff0<\/strong><\/p>\n<p><strong>3. \u8de8\u6a21\u6001\u6df1\u5ea6\u5b66\u4e60\u6280\u672f\u5728\u4e34\u5e8a\u5f71\u50cf\u4e2d\u7684\u5e94\u7528<\/strong><\/p>\n<p><strong>The Application of Deep Learning and Cross-modal Fusion Methods in Medical Imaging<\/strong><\/p>\n<p><strong>Abstract<\/strong><\/p>\n<p>Deep learning technology is gaining widespread prominence across various fields in this era. In the realm of medical imaging, it has steadily assumed a pivotal role in tasks such as feature recognition, object detection, and image segmentation, since its inception. With the continuous evolution of imaging techniques, the individual patient often possesses an expanding wealth of multi-modal imaging data. It is evident that deep learning models utilizing cross-modal image fusion techniques will find diverse applications in a lot more clinical scenarios. In the future, deep learning will play a significant role in the medical sector, encompassing screening, diagnosis, treatment, and long-term disease management. To provide a reference for future research, this review aims to present a concise overview of the fundamental principles of deep learning, the nature of cross-modal fusion methods based on deep learning, with their wide-ranging applications, and a comprehensive survey of the present clinical uses of single-modal and cross-modal deep learning techniques in medical imaging, with a particular emphasis on bone metastasis imaging.<\/p>\n<p><strong>Keywords: deep learning, cross-modal, tumor imaging, bong metastasis<\/strong><\/p>\n<p><strong>3.1 Preface<\/strong><\/p>\n<p>Nowadays, data are generated in massive quantities in the healthcare sector, from sources such as high-resolution medical imaging, biosensors with continuous output of physiologic metrics, genome sequencing, and electronic medical records. The limits on the analysis of such data by humans alone have clearly been exceeded, necessitating an increased reliance on machines. The use of artificial intelligence (AI), the deep-learning subtype in particular, has been enabled by the use of labeled big data, along with markedly enhanced computing power and cloud storage. One field that has attracted particular attention for the application of AI is radiology, as the cost of medical scans is declining, and the use of imaging studies is now at least as common as physical examinations, even surpassing the latter in surgical emergencies out of humanism and accuracy concerns. AI can greatly aid clinicians in diagnosis, especially in the interpretation of radiographic images, the accuracy of which heavily relies on the clinical experience and scrutiny of their interpreters, thus freeing clinicians to devote more of their attention to providing bedside healthcare. The radiologic screening and staging of tumors rely heavily on radiologists\u2019 subjective judgments. For some minuscule or ambivalent lesions, it is often difficult to arrive at a definitive diagnosis based solely on clinical experience. A case reported by Dong et al. proves the vulnerability of relying on error-prone human judgment[1 . AI methods mainly analyze medical images through image processing and deep learning techniques. As an assistant for clinicians, deep neural networks can be trained with large datasets of radiologic images or clinical information, automatically learning features key to the revelation of pathology or lesion localization. In addition to deep learning models based on images of a single modality, researchers have also proven the feasibility of integrating multi-modal medical imaging data in algorithms, with improved model robustness. A combination of feature representations from different imaging modalities can effectively improve the performance of tumor detection, classification, and segmentation. Artificially generating relatively scarce imaging data from more easily accessible radiographs by way of cross-modal image translation can not only aid diagnoses but also improve the performance of deep learning models.<\/p>\n<p>This article will briefly review the relevant background of deep neural networks (DNNs), as well as the up-to-date development of cross-modal fusion and image translation methods. An overview of the current clinical applications of single-modal and cross-modal deep learning in tumor imaging, especially in bone metastasis imaging, is also provided.<\/p>\n<p><strong>3.2. Deep Neural Network (DNN)<\/strong><\/p>\n<p>Traditional machine learning methods face limitations in handling data in its raw form, as creating a suitable internal representation or feature vector requires a meticulous feature extractor designed manually to convert raw data, such as image pixels. Only after then a classifier, every detail of which was manually set and adjusted, could detect or classify patterns in the input, and spell out its outcome. Because of the varying qualities of images, lots of intricate image enhancement or filtering algorithms, such as adaptive Gaussian learning, histogram equalization, and de-noising, are designed alone for the purpose of pre-processing images to be ready for the feature extractor. Another downside of conventional machine learning is that manually coded algorithms would only, with images with finer details or better contrast, allow for automatic execution of the thought processes that can best mimic, not surpass, that of a clinician. All the key features to be extracted and used in encoding the classifier were essentially the same set of \u201cinputs\u201d a clinician would use to make his or her judgment.<\/p>\n<p>In contrast, \u201crepresentation learning\u201d is a set of techniques utilized in deep learning, which enable machines to analyze raw data without manual feature engineering. This allows the system to automatically identify the relevant patterns or features needed for classification or detection. Pattern recognition using deep neural networks (DNNs) can help interpret medical scans, pathology slides, skin lesions, retinal images, electrocardiograms, endoscopy, faces, and vital signs. Deep learning algorithms employ multiple tiers of representation through the composition of nonlinear modules, which transform the input representation from one level to the next, beginning with the raw input and continuing to higher and more abstract levels. It is helpful to think of the entire network as, nonetheless, a \u201cfunction\u201d, that takes in a set of inputs and spills out an output, though with absurdly complicated parameters and transformations. Irrelevant variations or noise can be lost as stepping up towards the higher layers of representation that amplify only features important for discrimination. By this \u201clayering\u201d method, very complex functions can be learned. A key differentiating feature of deep learning compared with other subtypes of AI is its autodidactic quality, i.e. neither the number of layers nor features of each layer is designed by human engineers, unencumbered by either the essence or the flaws of the human brain.<\/p>\n<p><strong>3.2.1. Supervised learning<\/strong><\/p>\n<p>Supervised learning is the process of learning with training labels assigned by the supervisor, i.e. the training set of examples has its raw input data bundled with their desired outputs. When used to classify images into different categories, the machine is shown an image and produces an output in the form of a vector of scores, one for each category during training. In supervised learning, the machine receives immediate feedback on its performance when its output does not match the expected output. The aim is to assign the highest score to the desired category among all categories. This is achieved by calculating a cost function that measures the average error or distance between the output scores and the expected pattern of scores. The inputs of the cost function are the parameters of the machine. Much like the \u201cupdate rule\u201d used in the Mixed Complementarity Problem (MCP), the perceptron learning algorithm updates its internal adjustable parameters to minimize errors when it predicts the wrong category[2]. The adjustable parameters consist of weights and biases, where weights control the input-output function of the machine. The algorithm learns from its mistakes rather than successes, and the weights can number in the hundreds or millions. Weights are assigned to the connections between neurons from the input layer and one of the neurons in the next layer, in some sense representing the \u201cstrength\u201d of a connection. The activation of a single notch of neurons in the next layer was computed by taking the weighted sum of all the activations of the first layer, e.g., greyscale values of the pixels. Just like biological neurons may have different thresholds that the graded sum of electric potentials at the cell body needed to reach for axonal propagation, the algorithm may not want its neuron to light up simply over a sum greater than 0. So, a \u201cbias for inactivity\u201d is introduced into the weighted sum formula. For example, if a neuron is designed to be active only if the weighted sum exceeds 10, then a 10 is subtracted from the formula before the transformation that follows. Weights represent the pixel pattern (weight assigned to each pixel can be visualized as a pixel pattern) that the algorithm identifies, while biases provide a threshold indicating the required level of weighted sum for a neuron to become meaningfully active. When the goal is to have the value of activations of the next layer between 0 and 1 and were the mapping to be smooth and linear, then the weighted sum can be pumped into a sigmoid function, i.e. 1\/(1 + exp(\u2212w)), where w is the weighted sum. A sigmoid transformation compresses the continuum of real numbers, mapping them onto the interval between 0 and 1, effectively pushing negative inputs towards zero, and positive inputs towards 1, and the output steadily increases around the input 0. Say the mapping is from p neurons from the one layer to the q neurons of the next layer, there would be p\u00d7q number of weights and q biases. These are all adjustable parameters that can be manipulated to modify the behavior of this network. In a deep learning system, changing the parameters may reflect a shift in the location, size, or shape of the representations to find \u201cbetter\u201d features to travel through the layers to get to the desired output.<\/p>\n<p>At present though, preferred mappings in DNN are neither smooth nor linear. By applying a non-linear function to the input, the categories become separable by the last output layer in a linear way, resulting in a definitive category output, unlike the previously mentioned range of numerical values that would require arbitrary cut-off points to finalize the categorization. A sigmoid transformation was once popular in the era of Multilayer Perceptron, during which a machine was simply an executor of commands, and the feature detected by each layer was designed and programmed by human engineers, such that the final output, as a continuous variable, would be interpretable[3]. The rectified linear unit (ReLU) is currently the most widely used non-linear function, which introduces non-linearity into the network by setting all negative values to zero. This is in contrast to the smoother non-linearities, such as tanh(z) or 1\/(1 + exp(\u2212z)), used in previous decades.<br \/>\nReLU has proven to be a faster learner in deep networks compared to these other non-linear functions and allows for the training of deep supervised networks without the need for unsupervised pre-training[4].<\/p>\n<p>This would not work in DNN, as hidden layers would not be picking up edges and patterns based on our expectations. How the machine gets to the correct output is still an enigma, and its intelligence still awaits revelation.<\/p>\n<p>The essential of learning by neural networks is to minimize the cost function. It is important for this cost function to have a nice and smooth output so that the local minimum can be obtained by taking little steps downhill, rather than being either on or off in a binary way the way biological neurons are. To adjust the weight and bias values of the parameter vector in a high-dimensional space, the learning algorithm computes a gradient vector that specifies how much the error, or cost, would increase or decrease if each parameter were slightly modified. In mathematical terms, this is similar to taking derivatives of a function with respect to a variable to observe the trend of the function between the two infinitesimally close values of that variable. In multivariate calculus, the gradient of a function indicates the path of the steepest incline, pointing towards the direction in the input space where one should move to minimize the output of this cost function with the utmost speed, and the length of the vector indicates exactly how steep the steepest ascent is. The weight vector is modified by shifting it in the opposite direction of the gradient vector, and the size of the adjustments is proportional to the slope of the gradient vector. When the slope of the gradient vector approaches the minimum, the step size decreases to prevent overshooting. This is the so-called \u201cgradient descent\u201d that converges on some local minimum. Minimizing the cost function can guarantee better performance across all training samples. Viewed from a different perspective, the gradient vector of the cost function encodes the relative importance of weights and biases, which changes to which weights matter the most to minimize the cost. The magnitude of each component represents<br \/>\nhow sensitive the cost is to each weight and bias.<\/p>\n<p>In practice, most practitioners use a procedure called stochastic gradient descent (SGD). It involves randomly selecting a few input vectors as mini-batches, computing the corresponding outputs, errors, and the gradient descent step. The weights were adjusted accordingly. This process is repeated for many small subsets of examples from the training set until the average cost function stops decreasing. Each small subset of examples gives a noisy estimate of the average gradient over all examples, and thus the \u201cstochasticity\u201d. Despite its simplicity, SGD often achieves good results with far less computation time than more complex optimization techniques[5].<\/p>\n<p><strong>3.2.2. Backpropagation<\/strong><\/p>\n<p>Recursively adjusting the weights in proportion to the activation of the second-to-last layer, vise vera, or altering the biases to decrease the cost for a single training sample is a single round of digital learning. In a nutshell, the backpropagation procedure is an algorithm of computing the gradient descent efficiently. Calculating the gradient of a cost function with respect to the weights in a stacked multilayer module is a practical application of the chain rule of derivatives. A key insight is that the derivative of the cost function concerning the input can be obtained by reversing the order of the layers, working from the higher to the lower layers. The process of backpropagation entails computing gradients through all layers. From the uppermost layer where the network generates predictions down towards the lowermost layer where the external input is introduced. Once these gradients have been calculated, it is straightforward to determine the gradients with respect to the weights and biases of each module. After these gradients are computed, it becomes a straightforward task to derive the gradients with respect to the weights and biases of each module. The average of desired changes, obtained by traversing the backpropagation route for alternate training samples, was the optimal adjustment that parameters could make to make the model performs better in the training set.<\/p>\n<p>It was commonly thought that a simple gradient descent would get trapped in suboptimal local minima \u2014 weight configurations for which no small change would reduce the cost function, as finding the global minimum would be an intractable task. Recent theoretical and empirical results strongly suggest the cost function&#8217;s landscape is actually filled with a huge number of saddle points where the gradient is zero, indicating that the optimization challenge is more complex than originally thought, but most of these points have similar cost function values[6]. In other words, the depth of the local minima is almost the same across different saddle points, so it is not crucial which one the algorithm gets stuck at.<\/p>\n<p><strong>3.2.3. Convolutional neural networks (CNN)<\/strong><\/p>\n<p>Convolutional neural networks (CNNs) are easier to train and have better generalization capabilities compared to other feedforward networks with fully connected layers. They are specifically designed to process data represented as multiple arrays, such as grayscale images consisting of a single 2D array containing pixel intensities of varying values.<\/p>\n<p>The four key ideas behind CNN are inspired by the properties of natural signals and visual neuroscience: local connections, shared weights, pooling, and the use of multi-layer. The convolutional and pooling layers are directly inspired by the concept of simple cells and complex cells, respectively, in the visual cortex, and the overall architecture is reminiscent of the LGN-V1-V2-V4-IT hierarchy in the visual cortex&#8217;s ventral pathway[7][8]. Local groups of values in array data often exhibit high correlation and form characteristic local motifs that can be readily identified. Therefore, pattern recognition makes CNN most useful in detecting images.<\/p>\n<p><strong>3.2.3.1. Convolution<\/strong><\/p>\n<p>The main function of the convolutional layer is to identify and extract local combinations of features from the preceding layer (Fig. 1). The actual process of matching is accomplished through filtering in the convolutional layer: A filter bank can be thought of as a small matrix of representative features (of real numbers) for which the number of rows and columns, eg. n\u00d7 n, of pixels is arbitrarily set. The filter and image patch are lined up, and each image pixel is multiplied (dot product) by the corresponding feature. The result is added up and divided by the total number of pixels in the filter to arrive at a specific feature value. Feature value indicates how well the feature is represented at that position. Sliding over n pixels, the same procedure is repeated for every n\u00d7 n block of pixels for the entire input image, and a feature map, a \u201cmap\u201d of where the filter feature occurs, is obtained. All units in a feature map share the same filter bank. Therefore, the local characteristics of images and other signals remain constant regardless of their location. In simpler terms, a pattern that appears in one part of the image can appear in any other part as well. Hence, the approach of employing units with identical weights to identify corresponding patterns across various sections of the array is adopted. In a convolution layer, filtering can be performed for a bunch of features and create a stack of filtered images. Each feature map in a layer employs its own filter bank. From a mathematical standpoint, the operation of filtering executed by a feature map can be described as a discrete convolution, hence the name.<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-03-at-9.46.57\u202fAM.png\" alt=\"img\" \/><\/p>\n<p>Fig. 1: Example of a filter (kernel) convolution. Note the new pixel value shown in the figure has not been weighted by the number of windows in the filter.<\/p>\n<p><strong>3.2.3.2. Pooling<\/strong><\/p>\n<p>The aim of the pooling layer is to reduce the size of a feature map by merging similar features into a single one through the following steps: (1) choose an appropriate window size, usually 2\u00d72 or 3\u00d73 pixels; (2) pick a stride (by how many pixels the window steps down to run through a feature map) accordingly, usually 2 pixels; (3) walk the window by its stride across the filtered images; (4) take the maximum value in each window as the pooling result and form a \u201cpooled map\u201d. A robust motif detection can be accomplished by simplifying the positioning of each feature across all feature maps that are fed into this CNN layer. Pooling helped the algorithm to neglect where in each window the maximum value occurs, making it less sensitive to small shifts in position, either parallel or rotational, an image that strongly fits the filter will still get picked up.<\/p>\n<p><strong>3.2.3.3. Normalization<\/strong><\/p>\n<p>To keep the math from blowing up, a layer is then passed through a non-linearity such as a ReLU making negative values all 0. This procedure of nonlinear transformation is \u201cnormalization\u201d in deep learning terms.<\/p>\n<p>The CNN architecture involves stacking multiple stages of convolution, non-linearity (normalization), and pooling on top of each other, followed by a final fully connected layer (Fig. 2). Each layer&#8217;s filter banks in the convolutional layers and voting weights in the fully connected layer are learned through the backpropagation algorithm. In the fully connected layer, also known as the dense layer due to the fact that a large number of neurons are densely connected with each other, a list of feature values becomes a list of votes, when timed by relevant weights that map this layer to the output layer gives the final answer. It is worth noting that this list of votes in the fully connected layer looks a lot alike a list of feature values. Indeed, the output of this layer as intermediate categories can still feed into the input of the next layer, propagating the cycle instead of becoming the final votes.<\/p>\n<p><img decoding=\"async\" src=\"https:\/\/chinadigitaltimes.net\/chinese\/files\/2025\/05\/Screenshot-2025-05-03-at-9.48.16\u202fAM.png\" alt=\"img\" \/><\/p>\n<p>Fig. 2: Example of a CNN with various types of layers. The convolutional layer does not decrease the size, i.e., the number of pixels, of the input figure, rather it encodes the feature of its input. The pooling layer does decrease the size of its input. The amount of which the size decreases depends on the size of the pooling window and the stride.<\/p>\n<p><strong>3.3. Cross-modal fusion<\/strong><\/p>\n<p>Cross-modal fusion refers to the process of integrating data from different modalities. PET\/CT is a classic example of cross-modal fusion. CT is a type of imaging modality that provides high-resolution, cross-sectional images with excellent clarity and density resolution. PET, on the other hand, is a nuclear imaging technique that generates images showing the spatial distribution of positron-emitting radiopharmaceuticals within the body. With less precise structural details, PET images are well capable of displaying metabolic activity. PET\/CT fuses CT with PET, possessing information on both the anatomical details and the metabolic spectrum. As each different information stream possesses unique characteristics, single-modal data often do not contain all the effective features to produce accurate results, whether for data analysis or prediction tasks. Cross-modal deep learning models combine data from two or more different modalities, learning different feature representations from different modalities and facilitating communication and transformation among different information streams, to accomplish specific downstream tasks. This special type of deep learning can improve the accuracy of predictions and enhance the robustness of models.<\/p>\n<p><strong>3.3.1. Cross-modal fusion methods<\/strong><\/p>\n<p>Cross-modal fusion methods can be categorized into three types: early fusion, late fusion, and hybrid fusion. In early fusion, unimodal features are combined into a single representation before the feature extraction or modeling process[9] . After feature extraction or modeling is performed separately to reduce unimodal features, the outputs are integrated to learn concepts and obtain the final prediction in late fusion[10] . Whereas hybrid fusion combines early and late fusion methods, where fusion is performed at both the feature level<br \/>\nand the output layer [11].<\/p>\n<p>There are various methods of early fusion, including operating on elements at the same position in different modalities. For example, in the field of medical imaging, different imaging modalities can be fused into an integrated image. Nefian et al. proposed a cross-modal early fusion method that used both the factorial and the coupled hidden Markov model for audio-visual integration in speech recognition[12] . Early fusion was done by multiplying the corresponding elements of visual features that capture mouth deformation over consecutive frames and the vector representation, representing the frequency of audio observations, learned by long short-term memory neural networks. A dimensionality reduction was then done on the observation vectors obtained by concatenation of the audio and visual features. Indeed, early fusion methods are often simple in structure with low computational complexity. However, the resulting feature is often high in dimensions, which can impose a significant computational burden on the subsequent model if dimensionality reduction is not performed.<br \/>\nAs an example of late fusion, in 2014, Simonyan et al. proposed an architecture that separately inputs spatial and temporal recognition streams of videos, where the spatial stream recognizes actions from still video frames, whilst the temporal stream is in charge of action recognition from motion in the form of dense optical flow[13]. The learned feature outputs are combined by late fusion via either averaging or a linear support vector machine (SVM). As fusion significantly improves on both streams alone, the result proves the complementary nature of inputs spatial and temporal recognition streams and that cross-modal fusion indeed preserves more information of use in the algorithm. Late fusion does not explicitly consider the inter-modality correlation at the feature level, which may result in a lack of interaction among different modalities at the feature level. Consequently, the resulting feature representations after cross-modal fusion may not be rich enough, potentially limiting the effectiveness of the fusion approach.<\/p>\n<p>There is no one optimal solution for all, and the choice of fusion method should be case-by-case.<\/p>\n<p><strong>3.3.2. Cross-modal image translation<\/strong><\/p>\n<p>Cross-modal image translation has gradually matured in the field of computer vision. Given sufficient training data, deep learning models are capable of learning discriminative features from images of different modalities, and the process of image-to-image translation can be viewed as transforming one potential representation of a scene to another.<\/p>\n<p>In 2017, Isola et al. released a Pix2Pix software that is effective at various image translation tasks, such as synthesizing photos from label maps, reconstructing objects from edge maps, and colorizing images[14] . Conventional CNN learns to minimize a loss function that is arbitrary, the making of which takes a lot of manual effort. The Pix2Pix software adopts conditional Generative Adversarial Networks (GANs) that automatically learn the loss function to train the mapping from input to output images, besides learning the mapping itself, as a generic solution to pixel-to-pixel prediction. These networks solve the whole genre of problems that used to require very different loss functions. Conditional GANs differ from other formulation in that it treats output pixels as mutually dependent and thus learns a structured loss, which penalizes the joint configuration of the output. Pix2Pix has good performance on many image translation tasks, but its ability in generating high-resolution images is suboptimal. Wang et al. improved upon Pix2Pix by proposing a new image translation framework for synthesizing high-resolution photo-realistic images from semantic label maps using conditional GANs in 2018. Compared to Pix2Pix, this method has two main improvements: image translation at 2048 \u00d7 1024 resolution and semantic editing of images. To generate high-resolution images, their method uses a coarse-to-fine generator, which is composed of a local enhancer for fine high-resolution image conversion and a global generator for coarse low-resolution image conversion respectively, a multi-scale discriminator architecture, and a robust adversarial learning objective function.<\/p>\n<p>Additionally, it adds a low-dimensional feature channel to the input, which allows for the generation of diverse results images based on the same input label map. Zhu et al. proposed the BicycleGAN model combining both the conditional Variational Autoencoder GAN approach and the conditional Latent Regressor GAN approach, based on Pix2Pix in 2017[15] . BicycleGAN is a technique for multi-modal image translation that accomplishes not just the primary objective of mapping the input, together with the latent code to the output, but also concurrently learns an encoder that maps the output back to the latent space. The bijection between the output and the latent space prevents multiple distinct latent codes from producing the same output, also known as non-injective mapping. BicycleGAN allows the generator to model a distribution of high-dimensional output given different encoders, producing diverse and realistic results while remaining faithful to the input.<\/p>\n<p>To accurately transform specific objects among different modalities is the main challenge of cross-modal image translation. Most cross-modal image translation methods require paired data as the input, and due to the scarcity of the paired data, the translated images are often suboptimal or suffer from mode collapse, where the output only represents a limited number of real samples, etc. Therefore, how to achieve high-quality cross-modal image translation with a small amount of paired data is a valuable direction for research.<\/p>\n<p><strong>3.4. The application of cross-modal deep learning<\/strong><\/p>\n<p>AI is increasingly being studied in metastatic skeletal oncology imaging, and deep learning has been assessed for tasks such as detection, classification, segmentation, and prognosis. Zhao et al. developed a deep neural network-based model to detect bone metastasis on whole-body bone scan (WBS), irrespective of the primary malignancy[16 . Compared to experienced nuclear medicine physicians, the deep learning model not only had a time savings of 99.88% for the same workload, but it also had better diagnostic performance, with improved accuracy and sensitivity. To overcome the constraint of the time-consuming effort required for precise labeling of large datasets, Han et al. proposed a 2D CNN classifier-tandem architecture named GLUE, which integrates whole body and local patches for WBS of prostate cancer patients[17 . The 2D-CNN modeling is the best fit for planar nuclear medicine scans, provided there is a massive amount of training data available. The GLUE model had significantly higher AUCs than a whole-body-based 2D CNN model when the labeled dataset used for training was limited. Noguchi et al. developed a deep learning-based algorithm, with high lesion-based sensitivity and low false positives, to detect bone metastases in CT scans[18 . An observer study was also done to evaluate its clinical efficacy, which showed improved radiologists&#8217; performance when aided by the model, with higher sensitivity, by both lesion-based and case-based analyses, in less amount of interpretation time. Fan et al. used AdaBoost and Chan-Vese algorithms to detect and segment sites of spinal metastasis of lung cancer on MRI images[19 . Chan-Vese algorithm had the best performance. The accuracy of the segmentation, expressed in terms of DSC and Jaccard coefficient scores, were 0.8591 and 0.8002, respectively. Liu et al. built a deep learning model based on 3D U-Net algorithms for the automatic segmentation of pelvic bone and sites of prostate cancer metastases on MRI-DWI and T1-weighted MRI images[20 . The model was found to work best on patients with few metastases, boosting the use of CNN as an aid in M-staging in clinical practice. Multiple deep classifiers were developed by Lin et al. to automatically detect metastases in 251 thoracic SPECT bone images[21 . The performance of the classifiers was found to be excellent, with an AUC of 0.98. Moreau et al. compared different deep learning approaches to segment bones and metastatic lesions in PET\/CT images of breast cancer patients[22]. The results indicated that the U-NetBL-based approach for bone segmentation outperformed traditional methods, with a mean DSC of 0.94 \u00b1 0.03, whereas the traditional methods struggled to distinguish metabolically active organs from the bone draft.<\/p>\n<p>Compared to the aforementioned deep learning examples, the more avant-garde cross-modal image fusion and translation techniques have not been widely investigated in bone metastasis imaging. Xu et al. adopted two different convolutional neural networks for lesion segmentation and detection and combined the spatial feature representations extracted from the two different modalities of PET and CT[23] . Their cross-modal method completed the three-dimensional detection of multiple myeloma, outperforming traditional machine learning methods. The research conducted by Wang et al. revealed that texture features extracted from multiparametric prostate MRI before intervention, when combined with clinicopathological risks such as free PSA level, Gleason score, and age, could effectively predict bone metastasis in patients with prostate cancer[24] . The outcome of this study can be seen as a proof of concept for the significance of cross-modal data.<\/p>\n<p>Even though cross-modal investigations regarding the sites of bone metastases are limited by now, there has been plenty of evidence proving the utility of cross-modal fusion in oncological imaging. These applications and trains of thought can be well extrapolated to the field of osseous metastasis imaging. Cross-modal fusion can be applied to tasks such as tumor detection, segmentation, and classification to improve model the performance of deep learning models. Cross-modal image translation can be used for data augmentation to facilitate various downstream tasks.<\/p>\n<p>Cross-modal fusion methods are often employed to enrich the models with cross-modal image features, thus improving the performance of tumor detection. Further, convolutional neural networks are used to capture the relationships between adjacent pixels and extract effective features from the image in deep learning-based cross-modal tumor detection algorithms. In 2021, Huang et al. proposed a ResNet network-based framework, AW3M, that used ultrasonography of four different modalities jointly to diagnose breast cancer[25]. By combining the cross-modal data, the AW3M based upon multi-stream CNN equipped with self-supervised consistency loss was utilized to extract both modality-specific and modality-invariant features, with improved diagnostic performance.<\/p>\n<p>As for tumor segmentation, many researchers hinge on either the four types of MRI image modalities or the two modalities of PET\/CT encompassing anatomical and metabolic information to perform cross-modal fusion and improve segmentation performance. For instance, Ma et al. explored CNN-based cross-modal approaches for automated nasopharyngeal carcinoma segmentation[26]. Their proposed multi-modality CNN utilizes CT and MRI to jointly learn a cross-modal similarity metric and fuse complementary features at the output layer to segment paired CT-MR images, demonstrating exceptional performance. Additionally, the study combines the features extracted from each modality&#8217;s single-modality CNN and multi-modality CNN to create a combined CNN that capitalizes on the unique characteristics of each modality, thereby improving segmentation performance. Fu et al. introduced a deep learning-based framework for multimodal PET-CT segmentation that leverages PET&#8217;s high tumor sensitivity in 2021[27 . Their approach utilized a multimodal spatial attention module to highlight tumor regions and suppress normal regions with physiologic high uptake from PET input. The spatial attention maps generated by the PET-based module were then used to target a U-Net backbone for the segmentation of areas with higher tumor likelihood at different stages from CT images. Results showed that their method surpasses the state-of-the-art lung tumor segmentation approach by 7.6% in the Dice similarity coefficient.<\/p>\n<p>As the diagnostic procedure often requires the integration of multi-modal information, such as chief complaints, physical examinations, medical histories, laboratory tests, and radiology, cross-modal fusion methods are also commonly utilized in disease classification tasks. Cross-modal fusion synthesizes data from different modalities to enrich effective feature representations, enabling deep learning models to extract useful information from different modalities to aid in diagnosis. Zhang et al. proposed a technique for prostate cancer diagnosis using a multi-modal combination of B-mode ultrasonography and sonoelastography[28] . Quantitative features such as intensity statistics, regional percentile features, and texture features were extracted from both modalities, and an integrated deep network was proposed to learn and fuse these multimodal ultrasound imaging features. The final step of disease classification was completed by a support vector machine.<\/p>\n<p>Due to the relative scarcity of medical images, cross-modal image translation is often used to synthesize part of the data in the training set as a data augmentation method for a better-performing deep learning model with a small sample size. Since integrated data from different modalities often exhibit better performance in deep learning models, the multi-modal image data input generated by cross-modal image translation methods can be directly used as targets for tumor detection. A two-step approach for semi-supervised tumor segmentation using MRI and CT images was proposed by Jiang et al[29 . The first step is a tumor-aware unsupervised cross-modal adaptation using a target-specific loss to preserve tumors on synthesized MRIs from CT images. The second step involves training a U-Net model with synthesized and limited original MRIs using semi-supervised learning. Semi-supervised learning is used to boost the accuracy (80%) of tumor segmentation by combining labeled pre-treatment MRI scans with synthesized MRIs, while training with synthesized MRIs had an accuracy of 74%. The proposed approach demonstrated the effectiveness of tumor-aware adversarial cross-modal translation for accurate cancer segmentation from limited imaging data.<\/p>\n<p>In general, there have been bounties of research supporting the application of deep learning in bone metastasis imaging, but the specific application of cross-modal fusion methods is still lacking. Whereas, clinical evaluations regarding bone metastasis often require multi-modal data, such as a chief complaint of lower back pain, a past medical history of pathological fractures, a positive genetic test for specific mutations indicating a higher risk of bone metastasis, or increased blood calcium and alkaline phosphatase concentrations in laboratory reports, etc. Therefore, evaluating osseous lesions with multi-modal data can improve the specificity of diagnosis and reduce the false positive rates in the diagnostic and treatment process. The application of cross-modal deep learning methods in the field of bone metastasis imaging and diagnosis is worth further exploration.<\/p>\n<p><strong>3.5. conclusion<\/strong><\/p>\n<p>The above review covers the definition and basic principles of deep learning and cross-modal image generation and fusion methods, briefly describes some common cross-modal deep learning algorithms, and summarizes bits of current research on the application of deep learning models in medical imaging, especially bone metastasis imaging. Compared to traditional deep learning models fed with data input of a single modality, multi-modal methods are more recent, with a limited number of relevant research. Given the increasing prevalence of cancer screening and the significant surge in patient-specific clinical data, including radiographs and laboratory tests, it is reasonable to anticipate an unparalleled demand for advanced, intelligent cross-modal deep learning methods in the future. Nevertheless, the use of AI in medical imaging analysis faces various challenges and limitations. These include the need for extensive and diverse datasets for training and validation, the potential for bias and overfitting, as well as the inherent black-box nature of deep learning algorithms[30 . Even though the demand for a large training set reiterates the merit of cross-modal deep learning, which enables the automatic generation of sample images through cross-modal image translation, the size of the training set still has a profound impact on the performance of algorithms. In parallel, the demand for \u201cexplainability\u201d has led to the notion of \u201cinterpretable machine learning\u201d, utilizing heat maps and metrics to track the focus of deep neural networks[31] . Overall, there is still much to be investigated regarding the application of cross-modal deep learning in the field of medical imaging.<\/p>\n<p>In summary, the project should be founded on the application of cross-modal deep learning techniques to offer practical solutions for challenges encountered in the clinical setting.<\/p>\n<p><strong>\u53c2\u8003\u6587\u732e<\/strong><\/p>\n[1] Dong X, Wu D. A rare cause of peri-esophageal cystic lesion[J] . Gastroenterology, 2023, 164(2): 191-193.<br \/>\n[2] Aswathi R R, Jency J, Ramakrishnan B, et al. Classification Based Neural Network Perceptron Modelling with Continuous and Sequential data[J] . Microprocessors and Microsystems, 2022: 104601.<br \/>\n[3] Gardner M W, Dorling S R. Artificial neural networks (the multilayer perceptron)\u2014a review of applications in the atmospheric sciences[J] . Atmospheric environment, 1998, 32(14-15): 2627-2636.<br \/>\n[4] Glorot X, Bordes A, Bengio Y. Deep Sparse Rectifier Neural Networks[J] . Journal of Machine Learning Research, 2011, 15: 315-323.<br \/>\n[5] Bottou L, Bousquet O. The tradeoffs of large scale learning[J] . Advances in Neural Information Processing Systems, 2007, 20: 1-8.<br \/>\n[6] Dauphin, Y. et al. Identifying and attacking the saddle point problem in high-dimensional non-convex optimization[J] . Advances in neural information processing systems, 2014, 27: 2933\u20132941.<br \/>\n[7] Hubel D H, Wiesel T N. Receptive fields, binocular interaction and functional architecture in the cat&#8217;s visual cortex[J] . Journal of Physiology, 1962, 160(1): 106-154.<br \/>\n[8] Cadieu C F, Hong H, Yamins D, et al. Deep neural networks rival the representation of primate IT cortex for core visual object recognition[J] . Plos Computational Biology, 2014, 10(12): e1003963.<br \/>\n[9] Nefian A V, Liang L, Pi X, et al. Dynamic bayesian networks for audio-visual speech recognition[J] . EURASIP Journal on Advances in Signal Processing, 2002, 2002(11):1-15.<br \/>\n[10] Snoek C G M, Worring M, Smeulders A W M. Early versus late fusion in semantic video analysis[C] . Proceedings of the 13th annual ACM international conference on Multimedia, 2005: 399-402.<br \/>\n[11] Wu Z, Cai L, Meng H. Multi-level fusion of audio and visual features for speaker identification[C] . International Conference on Biometrics Springer, Berlin, Heidelberg, 2005: 493-499.<br \/>\n[12] Nefian A V, Liang L, Pi X, et al. Dynamic Bayesian networks for audio-visual speech recognition[J] . EURASIP Journal on Advances in Signal Processing, 2002, 2002(11):1-15.<br \/>\n[13] Simonyan K, Zisserman A. Two-stream convolutional networks for action recognition in videos[J] . Advances in neural information processing systems, 2014, 27: 568\u2013576.<br \/>\n[14] Isola P, Zhu J Y, Zhou T, et al. Image-to-image translation with conditional adversarial networks[C] . Proceedings of the IEEE conference on computer vision and pattern recognition, 2017: 1125-1134.<br \/>\n[15] Zhu J Y, Zhang R, Pathak D, et al. Toward multimodal image-to-image translation[J]. Advances in neural information processing systems, 2017, 30: 465\u2013476.<br \/>\n[16] Zhao Z, Pi Y, Jiang L, Xiang Y, Wei J, Yang P, et al. Deep neural network based artificial intelligence assisted diagnosis of bone scintigraphy for cancer bone metastasis[J] . Scientific Reports, 2020, 10(1): 17046.<br \/>\n[17] Han S, Oh J S, Lee J J. Diagnostic performance of deep learning models for detecting bone metastasis on whole-body bone scan in prostate cancer[J] . European journal of nuclear medicine and molecular imaging, 2021, 49(2): 1-11.<br \/>\n[18] Noguchi S, Nishio M, Sakamoto R, Yakami M, Fujimoto K, Emoto Y, et al. Deep learning\u2013based algorithm improved radiologists\u2019 performance in bone metastases detection on CT[J] . European Radiology, 2022, 32(11): 7976-7987.<br \/>\n[19] Fan X, Zhang X, Zhang Z, Jiang Y. Deep learning on MRI images for diagnosis of lung cancer spinal bone metastasis[J] . Contrast Media &amp; Molecular Imaging, 2021, 2021(1): 1-9.<br \/>\n[20] Liu X, Han C, Cui Y, Xie T, Zhang X, Wang X. Detection and segmentation of pelvic bones metastases in MRI images for patients with prostate cancer based on deep learning[J] . Frontiers in Oncology, 2021, 11: 773299.<br \/>\n[21] Lin Q, Li T, Cao C, Cao Y, Man Z, Wang H. Deep learning based automated diagnosis of bone metastases with SPECT thoracic bone images[J] . Scientific Reports, 2021, 11(1): 4223.<br \/>\n[22] Moreau N, Rousseau C, Fourcade C, Santini G, Ferrer L, Lacombe M, et al. Deep learning approaches for bone and bone lesion segmentation on 18FDG PET\/CT imaging in the context of metastatic breast cancer[J] . 42nd Annual International Conference of the IEEE Engineering in Medicine &amp; Biology Society, 2020: 1532-1535.<br \/>\n[23] Xu L, Tetteh G, Lipkova J, et al. Automated whole-body bone lesion detection for multiple myeloma on 68Ga-pentixafor PET\/CT imaging using deep learning methods[J] . Contrast media &amp; molecular imaging, 2018, 2018: 2391925.<br \/>\n[24] Wang Y, Yu B, Zhong F, Guo Q, Li K, Hou Y, et al. MRI-based texture analysis of the primary tumor for pre-treatment prediction of bone metastases in prostate cancer[J] . Magnetic Resonance Imaging, 2019, 60: 76-84.<br \/>\n[25] Huang R, Lin Z, Dou H, et al. AW3M: An auto-weighting and recovery framework for breast cancer diagnosis using multi-modal ultrasound[J] . Medical Image Analysis, 2021, 72: 102137.<br \/>\n[26] Ma Z, Zhou S, Wu X, et al. Nasopharyngeal carcinoma segmentation based on enhanced convolutional neural networks using multi-modal metric learning[J] . Physics in Medicine &amp; Biology, 2019, 64(2): 025005.<br \/>\n[27] Fu X, Bi L, Kumar A, et al. Multimodal spatial attention module for targeting multimodal PET-CT lung tumor segmentation[J] . IEEE Journal of Biomedical and Health Informatics, 2021, 25(9): 3507-3516.<br \/>\n[28] Zhang Q, Xiong J, Cai Y, et al. Multimodal feature learning and fusion on B-mode ultrasonography and sonoelastography using point-wise gated deep networks for prostate cancer diagnosis[J] . Biomedical Engineering\/Biomedizinische Technik, 2020, 65(1): 87-98.<br \/>\n[29] Jiang J, Hu Y C, Tyagi N, et al. Tumor-aware, adversarial domain adaptation from CT to MRI for lung cancer segmentation[C] . International conference on medical image computing and computer-assisted intervention. Springer, Cham, 2018: 777-785.<br \/>\n[30] Castelvecchi D. Can we open the black box of AI?[J] . Nature, 2016, 538(7623): 20.<br \/>\n[31] Kuang C. Can A.I. Be Taught to Explain Itself?[J] . The New York Times, 2017, 21.<\/p>\n<p><strong>\u81f4\u8c22<\/strong><\/p>\n<p>\u9996\u5148\uff0c\u611f\u8c22\u6211\u7684\u5bfc\u5e08\uff0c\u90b1\u8d35\u5174\u9662\u58eb\uff0c\u6d93\u6d93\u5e08\u6069\uff0c\u94ed\u8bb0\u4e8e\u5fc3\u3002\u611f\u8c22\u5434\u5357\u5e08\u5144\uff0c\u77e5\u9047\u4e4b\u6069\u65e0\u4ee5\u62a5\u3002\u611f\u8c22\u5434\u4e1c\u8001\u5e08\uff0c\u4e09\u751f\u6709\u5e78\uff0c\u5f97\u60a8\u4f34\u6211\u4e00\u7a0b\u98ce\u96ea\u3002\u4e5f\u611f\u8c22\u6240\u6709\u53c2\u4e0e\u6b64\u9879\u76ee\u7684\u79d1\u7814\u5408\u4f5c\u4f19\u4f34\u548c\u8bfe\u9898\u7ec4\u7684\u5e08\u5144\u5e08\u5f1f\u4eec\uff0c\u4f60\u4eec\u7684\u5e2e\u52a9\u8ba9\u8fd9\u9879\u7814\u7a76\u5f97\u4ee5\u987a\u5229\u8fdb\u884c\u3002<\/p>\n<p>\u611f\u8c22\u5728\u534f\u548c\u9047\u5230\u7684\u6240\u6709\u8001\u5e08\u4eec\uff0c\u5b66\u751f\u673d\u6728\uff0c\u5e0c\u671b\u672a\u6765\u4e5f\u80fd\u5982\u4f60\u4eec\u4e00\u6837\uff0c\u4e0d\u8d1f\u4e00\u88ad\u767d\u8863\u3002<\/p>\n<p>\u6700\u540e\uff0c\u8981\u611f\u8c22\u6211\u7684\u5bb6\u4eba\u3002Wherever I go, this family is my fortress.<\/p>\n<p>\u8fd9\u8def\u9065\u9a6c\u6025\u7684\u4eba\u95f4\uff0c\u4f60\u6211\u5e73\u5b89\u559c\u4e50\u5c31\u597d\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u201c\u6700\u540e\uff0c\u8981\u611f\u8c22\u6211\u7684\u5bb6\u4eba\u3002Wherever I go, this family is my fortress. \u8fd9\u8def\u9065\u9a6c\u6025\u7684\u4eba\u95f4\uff0c\u4f60\u6211\u5e73\u5b89\u559c\u4e50\u5c31\u597d\u3002\u201d<\/p>\n<p>\u65e5\u524d\uff0c\u4e2d\u65e5\u53cb\u597d\u533b\u9662\u533b\u751f\u8096\u98de\u88ab\u5176\u59bb\u4e3e\u62a5\u5a5a\u5916\u60c5\uff0c\u5176\u4e2d\u4e00\u4f4d\u5f53\u4e8b\u4eba\u4e3a\u89c4\u57f9\u533b\u5e08\u3001\u534f\u548c\u533b\u5b66\u9662\u535a\u58eb\u751f\u8463\u88ad\u83b9\u3002\u4e3e\u62a5\u4fe1\u4e2d\u8fd8\u6307\u51fa\uff0c\u8096\u98de\u66fe\u5728\u624b\u672f\u8fc7\u7a0b\u4e2d\u4e0e\u62a4\u58eb\u53d1\u751f\u51b2\u7a81\uff0c\u671f\u95f4\u4e0d\u987e\u60a3\u8005\u5b89\u5371\uff0c\u4e0e\u8463\u88ad\u83b9\u4e00\u540c\u79bb\u5f00\u624b\u672f\u5ba4\u3002\u968f\u540e\uff0c\u8463\u88ad\u83b9\u7684\u6559\u80b2\u80cc\u666f\u5f15\u53d1\u7f51\u7edc\u5173\u6ce8\uff0c\u4e2d\u56fd\u77e5\u7f51\u5df2\u5c06\u5176\u535a\u58eb\u8bba\u6587\u4e0b\u67b6\u3002<\/p>\n","protected":false},"author":1061,"featured_media":717774,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_et_pb_use_builder":"","_et_pb_old_content":"","_et_gb_content_width":"","ampforwp-amp-on-off":"default","home_featured":"top","footnotes":"","_links_to":"","_links_to_target":""},"categories":[38917,45870,41492,45445,45438,45443,45435,43113],"tags":[50519,50528,42497,50228,50508],"class_list":["post-717840","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-38917","category-newsletter","category-twitter","category-social-media","category-citizen","category-platforms","category-cds-archives","category-43113","tag-50519","tag-50528","tag-42497","tag-50228","tag-50508","et-has-post-format-content","et_post_format-et-post-format-standard"],"_links":{"self":[{"href":"https:\/\/chinadigitaltimes.net\/chinese\/wp-json\/wp\/v2\/posts\/717840","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/chinadigitaltimes.net\/chinese\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/chinadigitaltimes.net\/chinese\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/chinadigitaltimes.net\/chinese\/wp-json\/wp\/v2\/users\/1061"}],"replies":[{"embeddable":true,"href":"https:\/\/chinadigitaltimes.net\/chinese\/wp-json\/wp\/v2\/comments?post=717840"}],"version-history":[{"count":0,"href":"https:\/\/chinadigitaltimes.net\/chinese\/wp-json\/wp\/v2\/posts\/717840\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/chinadigitaltimes.net\/chinese\/wp-json\/wp\/v2\/media\/717774"}],"wp:attachment":[{"href":"https:\/\/chinadigitaltimes.net\/chinese\/wp-json\/wp\/v2\/media?parent=717840"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/chinadigitaltimes.net\/chinese\/wp-json\/wp\/v2\/categories?post=717840"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/chinadigitaltimes.net\/chinese\/wp-json\/wp\/v2\/tags?post=717840"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}