SurgVisDom

此挑战的数据集已与挑战出版物一起公开发布。有关更多详细信息，请在此处查看synapse网站链接
外科数据科学正在彻底改变微创手术。通过开发具有上下文感知的算法，增加外科医生的令人兴奋的应用程序成为可能。但是，开发上下文感知模型所需的外科手术数据 (或更普遍的健康数据) 存在许多敏感性。这项挑战旨在探索外科手术中视觉领域适应的潜力，以克服数据隐私问题。特别是，我们建议使用来自类似临床任务的虚拟现实仿真数据的视频来开发算法来识别活动，然后在临床环境 (即猪模型) 中对相同任务的视频进行测试。

The dataset for this challenge has been publicly released alongside the challenge publication. For more details, please visit the Synapse website link here. Surgical data science is revolutionizing minimally invasive surgery. Developing context-aware algorithms enables exciting applications that augment surgeons' capabilities. However, surgical data (or more broadly, health data) required for developing context-aware models carries numerous sensitivity-related concerns. This challenge aims to explore the potential of visual domain adaptation in surgery to overcome data privacy issues. Specifically, we propose using videos from virtual reality simulation data of similar clinical tasks to develop activity recognition algorithms, which are then tested on videos of the same tasks conducted in a clinical setting (i.e., porcine models).

搜集汇总

背景与挑战

背景概述

SurgVisDom是一个用于外科手术视觉领域适应的数据集，通过虚拟现实仿真数据开发活动识别算法，并在猪模型临床视频中测试。该数据集由伦敦大学学院等机构于2020年发布，旨在克服健康数据隐私挑战。