面向开放缝合技能的视觉评估挑战——2024-2025 年 OSS 挑战赛
计算机视觉与模式识别
2026-05-22 v1 人工智能
机器学习
摘要
通过有效训练实现高水平手术技能对于获得最佳患者结果至关重要。自动化、数据驱动的技能评估具有显著潜力来提高手术培训。虽然机器学习方法日益受欢迎用于评估 minimally invasive surgery 中的技能，但其在开放手术中的应用仍有限。我们 presenting 一项专门用于基准和推动开放手术视觉技能评估的 MICCAI 挑战赛结果。挑战数据集包括使用固定 GoPro 摄像机在干燥实验室环境中记录的开放缝合训练任务视频，除主要视频模态外，还提供仪器轨迹。OSS 挑战赛在连续两年内举办，分别包括两项和三项独立任务：(1) 将技能水平分为四个类别进行分类，(2) 预测完整的 Objective Structured Assessment of Technical Skills（OSATS）在八个类别中的得分，(3) 跟踪手和外科工具。参赛者提交了包括深度学习视频模型、跟踪驱动方法和混合方法在内的多样化解决方案。通用时空视频模型始终实现了最强的性能，尽管概念上多样的方法在执行得当时也能达到竞争水平。预测细粒度 OSATS 分数仍具有挑战性，但通过增加训练数据可显著获益。关键点跟踪在频繁的遮挡和超出画面实例情况下仍具有挑战，限制了基于运动的技能分析的当前适用性。这项工作为手术技能评估提供了创新和多样化解决方案的基准，凸显了视频在开放手术中进行技能评估的前景及当前局限性，并确定了推动自动化技能评估迈向临床应用的关键方向。
关键词
引用
@article{arxiv.2605.22200,
title = {OSS: Open Suturing Skills Vision-Based Assessment Challenge 2024-2025},
author = {Hanna Hoffmann and Setareh Bady and Claas de Boer and Max Kirchner and Jan Egger and Rainer Röhrig and Frank Hölzle and Lennart Johannes Gruber and Kunpeng Xie and Marlon Neuhaus and Victor Alves and Guilherme Barbosa and Leonardo Barroso and João Carvalho and Hao Chen and Gabriella d'Albenzio and André Ferreira and Nuno Gomes and Yuichiro Hayashi and Kousuke Hirasawa and Rebecca Hisey and Seungjae Hong and Seoi Jeong and Tiago Jesus and Daehong Kang and Satoshi Kasai and Shunsuke Kikuchi and Takayuki Kitasaka and Satoshi Kondo and Hyoun-Joong Kong and Youngbin Kong and Atsushi Kouno and Shlomi Laufer and Kyu Eun Lee and Bining Long and Nooshin Maghsoodi and Hiroki Matsuzaki and Evangelos Mazomenos and Ori Meiraz and Kensaku Mori and Marina Music and Masahiro Oda and Roi Papo and Jieun Park and Rafael Piexoto and Saeid Rezaei and Mariana Ribeiro and Soyeon Shin and Yang Shu and Idan Smoller and Danail Stoyanov and Yihui Wang and Xinkai Zhao and Sebastian Bodenstedt and Isabel Funke and Stefanie Speidel and Behrus Hinrichs-Puladi},
journal= {arXiv preprint arXiv:2605.22200},
year = {2026}
}
备注
Stefanie Speidel and Behrus Hinrichs-Puladi jointly supervised this work. Submitted to MEDIA