\u30ed\u30dc\u30c3\u30c8\u652f\u63f4\u524d\u7acb\u817a\u5168\u6458\u8853\u306b\u304a\u3051\u308b\u795e\u7d4c\u8840\u7ba1\u675f\uff08NVB\uff09\u6e29\u5b58\u4e88\u6e2c\u3092\u76ee\u7684\u306b\u3001\u81ea\u5df1\u6559\u5e2b\u3042\u308a\u5b66\u7fd2\u3068\u591a\u30bf\u30b9\u30af\u5b66\u7fd2\u3092\u7d71\u5408\u3057\u305fHybrid Teacher\u2013Student\u30e2\u30c7\u30eb\u3092\u63d0\u6848\u3057\u305f\u3002<\/p>\n\n\n\n
Diego Andr\u00e9s delAguilaMoraga1
\u00b7 Huitao Wang1 \u00b7Satoshi Ando2 \u00b7Hayato Hoshina2 \u00b7Hiroshi Kawahira3 \u00b7
Yukihiro Nomura4 \u00b7Toshiya Nakaguchi4
Received: 9 April 2025 \/ Accepted: 30 October 2025
\u00a9TheAuthor(s) 2026
Abstract
Purpose Preserving neurovascular bundles (NVB) during robot-assisted radical prostatectomy (RARP) is vital for reducing
postoperativecomplicationssuchasurinaryincontinenceanderectiledysfunction.Buildingonourpreviousworkinensemble
based NVB classification, we propose the hybrid self-supervised teacher\u2013student model (Hybrid T\u2013S model) that leverages
multi-task learning to predict NVB preservation in prostatectomy videos.
Methods Our approach integrates a self-supervised framework (DINO) as an online self-distillation objective on multi-crop
views to learn robust embeddings in a limited data setting, rather than as a stand-alone large-scale pretraining. A teacher
encoder, which is an exponential moving average (EMA) of the student encoder, and a reconstruction decoder are trained
jointly with a classification head in a single end-to-end framework. This model is evaluated on single frames from patients
who underwent RARP surgery.
Results Our experimental evaluation shows that the Hybrid T\u2013S model outperforms previous NVB classification methods.
This highlights the benefits of integrating self-supervised learning and multi-task objectives in this surgical context. We
achieved an average accuracy of 86.55%, precision of 83.93%, recall of 90.73%, F1-score of 87%, and AUROC of 88.35%,
based on fivefold cross-validation.
Conclusion Incorporating representation learning through self-distillation, classification, and reconstruction provides com
plementary signals that enhance the prediction of NVB preservation. Our Hybrid T\u2013S model can assist surgeons in real
decision-making and improve patient recovery.
Keywords Self-supervised learning \u00b7 Multi-task learning \u00b7 Image classification \u00b7 RARP \u00b7 NVB
Introduction
BDiegoAndr\u00e9s del Aguila Moraga
andresdelaguila.moraga@gmail.com
Huitao Wang
wanghuitao@chiba-u.jp
Satoshi Ando
ando@jichi.ac.jp
Hayato Hoshina
bm3pt9@gmail.com
Hiroshi Kawahira
kawahira@jichi.ac.jp
Yukihiro Nomura
ynomura@chiba-u.jp
Toshiya Nakaguchi
nakaguchi@faculty.chiba-u.jp
<\/p>\n\n\n\n
\u30ed\u30dc\u30c3\u30c8\u652f\u63f4\u524d\u7acb\u817a\u5168\u6458\u8853\uff08robot-assisted radical prostatectomy\uff1aRARP\uff09\u306b\u304a\u3044\u3066\u3001\u8853\u5f8c\u306e\u5c3f\u5931\u7981\u3084\u52c3\u8d77\u969c\u5bb3\u3068\u3044\u3063\u305f\u5408\u4f75\u75c7\u3092\u8efd\u6e1b\u3059\u308b\u305f\u3081\u306b\u795e\u7d4c\u8840\u7ba1\u675f\uff08neurovascular bundle\uff1aNVB\uff09\u306e\u6e29\u5b58\u304c\u91cd\u8981\u3067\u3042\u308b\u3002\u624b\u8853\u52d5\u753b\u304b\u3089\u9ad8\u7cbe\u5ea6\u306b\u4e88\u6e2c\u53ef\u80fd\u306a\u81ea\u5df1\u6559\u5e2b\u3042\u308a\u5b66\u7fd2\u3068\u591a\u30bf\u30b9\u30af\u5b66\u7fd2\u3092\u7d71\u5408\u3057\u305f\u65b0\u305f\u306aHybrid Teacher\u2013Student\uff08Hybrid T\u2013S\uff09\u30e2\u30c7\u30eb\u3092\u63d0\u6848\u3057\u305f\u3002\u672c\u624b\u6cd5\u306e\u7279\u5fb4\u306f\u3001\u81ea\u5df1\u6559\u5e2b\u3042\u308a\u5b66\u7fd2\u30d5\u30ec\u30fc\u30e0\u30ef\u30fc\u30af\u3067\u3042\u308bDINO\u3092\u3001\u5927\u898f\u6a21\u4e8b\u524d\u5b66\u7fd2\u3068\u3057\u3066\u3067\u306f\u306a\u304f\u3001\u9650\u3089\u308c\u305f\u533b\u7642\u30c7\u30fc\u30bf\u74b0\u5883\u4e0b\u306b\u304a\u3051\u308b\u30aa\u30f3\u30e9\u30a4\u30f3\u81ea\u5df1\u84b8\u7559(self-distillation)\u3068\u3057\u3066\u6d3b\u7528\u3057\u3066\u3044\u308b\u70b9\u306b\u3042\u308b\u3002\u5b66\u751f\u30e2\u30c7\u30eb\u3068\u3001\u305d\u306e\u6307\u6570\u79fb\u52d5\u5e73\u5747\u306b\u3088\u308a\u66f4\u65b0\u3055\u308c\u308b\u6559\u5e2b\u30e2\u30c7\u30eb\u3092\u7528\u3044\u3001\u5206\u985e\u30bf\u30b9\u30af\u306b\u52a0\u3048\u3066\u518d\u69cb\u6210\u30bf\u30b9\u30af\u3092\u540c\u6642\u306b\u5b66\u7fd2\u3059\u308b\u30a8\u30f3\u30c9\u30c4\u30fc\u30a8\u30f3\u30c9\u69cb\u6210\u3068\u3059\u308b\u3053\u3068\u3067\u3001\u3088\u308a\u9811\u5065\u3067\u6c4e\u5316\u6027\u80fd\u306e\u9ad8\u3044\u7279\u5fb4\u8868\u73fe\u306e\u7372\u5f97\u3092\u53ef\u80fd\u306b\u3057\u305f\u3002RARP\u3092\u53d7\u3051\u305f\u60a3\u8005\u306e\u624b\u8853\u52d5\u753b\u304b\u3089\u62bd\u51fa\u3057\u305f\u5358\u4e00\u30d5\u30ec\u30fc\u30e0\u3092\u7528\u3044\u305f\u8a55\u4fa1\u306e\u7d50\u679c\u3001\u672c\u30e2\u30c7\u30eb\u306f\u5f93\u6765\u306eNVB\u5206\u985e\u624b\u6cd5\u3092\u4e0a\u56de\u308b\u6027\u80fd\u3092\u793a\u3057\u30015\u5206\u5272\u4ea4\u5dee\u691c\u8a3c\u306b\u304a\u3044\u3066\u5e73\u5747\u6b63\u89e3\u738786.55%\u3001F1\u30b9\u30b3\u30a287%\u3001AUROC88.35%\u3092\u9054\u6210\u3057\u305f\u3002\u3053\u308c\u3089\u306e\u7d50\u679c\u304b\u3089\u3001\u81ea\u5df1\u84b8\u7559\u3001\u5206\u985e\u3001\u518d\u69cb\u6210\u3068\u3044\u3046\u8907\u6570\u306e\u5b66\u7fd2\u4fe1\u53f7\u3092\u7d44\u307f\u5408\u308f\u305b\u308b\u3053\u3068\u304cNVB\u6e29\u5b58\u4e88\u6e2c\u306b\u6709\u52b9\u3067\u3042\u308b\u3053\u3068\u304c\u793a\u3055\u308c\u305f\u3002\u672c\u7814\u7a76\u3067\u63d0\u6848\u3059\u308bHybrid T\u2013S\u30e2\u30c7\u30eb\u306f\u3001\u8853\u4e2d\u306e\u610f\u601d\u6c7a\u5b9a\u652f\u63f4\u3084\u5916\u79d1\u533b\u306e\u5224\u65ad\u88dc\u52a9\u306b\u8ca2\u732e\u3057\u3001\u60a3\u8005\u306e\u8853\u5f8c\u6a5f\u80fd\u56de\u5fa9\u306e\u5411\u4e0a\u306b\u3064\u306a\u304c\u308b\u53ef\u80fd\u6027\u304c\u3042\u308b\u3002<\/p>\n\n\n\n
\u30ad\u30fc\u30ef\u30fc\u30c9\uff1a<\/strong>\u30ed\u30dc\u30c3\u30c8\u652f\u63f4\u4e0b\u624b\u8853\u3001\u524d\u7acb\u817a\u3001\u795e\u7d4c\u6e29\u5b58\u3001\u6a5f\u68b0\u5b66\u7fd2\u3001\u624b\u8853\u52d5\u753b\u89e3\u6790<\/strong><\/strong><\/p>\n","protected":false},"excerpt":{"rendered":" \u30ed\u30dc\u30c3\u30c8\u652f\u63f4\u524d\u7acb\u817a\u5168\u6458\u8853\u306b\u304a\u3051\u308b\u795e\u7d4c\u8840\u7ba1\u675f\uff08NVB\uff09\u6e29\u5b58\u4e88\u6e2c\u3092\u76ee\u7684\u306b\u3001\u81ea\u5df1\u6559\u5e2b\u3042\u308a\u5b66\u7fd2\u3068\u591a\u30bf\u30b9\u30af\u5b66\u7fd2\u3092\u7d71\u5408\u3057\u305fHybrid Teacher\u2013Student\u30e2\u30c7\u30eb\u3092\u63d0\u6848\u3057\u305f\u3002 Diego Andr\u00e9s delAguilaMoraga1\u00b7 Huitao Wang1 \u00b7Satoshi Ando2 \u00b7Hayato Hoshina2 \u00b7Hiroshi Kawahira3 \u00b7Yukihiro Nomura4 ...\u7d9a\u304d\u3092\u8aad\u3080<\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[67],"tags":[],"class_list":["post-13398","post","type-post","status-publish","format-standard","hentry","category-edu-res"],"acf":[],"_links":{"self":[{"href":"https:\/\/www.jichi.ac.jp\/msc\/wordpress\/wp-json\/wp\/v2\/posts\/13398","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.jichi.ac.jp\/msc\/wordpress\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.jichi.ac.jp\/msc\/wordpress\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.jichi.ac.jp\/msc\/wordpress\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.jichi.ac.jp\/msc\/wordpress\/wp-json\/wp\/v2\/comments?post=13398"}],"version-history":[{"count":1,"href":"https:\/\/www.jichi.ac.jp\/msc\/wordpress\/wp-json\/wp\/v2\/posts\/13398\/revisions"}],"predecessor-version":[{"id":13399,"href":"https:\/\/www.jichi.ac.jp\/msc\/wordpress\/wp-json\/wp\/v2\/posts\/13398\/revisions\/13399"}],"wp:attachment":[{"href":"https:\/\/www.jichi.ac.jp\/msc\/wordpress\/wp-json\/wp\/v2\/media?parent=13398"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.jichi.ac.jp\/msc\/wordpress\/wp-json\/wp\/v2\/categories?post=13398"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.jichi.ac.jp\/msc\/wordpress\/wp-json\/wp\/v2\/tags?post=13398"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}