This website requires JavaScript.

Federated Cycling (FedCy): Semi-supervised Federated Learning of Surgical Phases

Hasan KassemDeepak AlapattPietro MascagniAI4SafeChole ConsortiumAlexandros KarargyrisNicolas Padoy
摘要
Recent advancements in deep learning methods bring computer-assistance a stepcloser to fulfilling promises of safer surgical procedures. However, thegeneralizability of such methods is often dependent on training on diversedatasets from multiple medical institutions, which is a restrictive requirementconsidering the sensitive nature of medical data. Recently proposedcollaborative learning methods such as Federated Learning (FL) allow fortraining on remote datasets without the need to explicitly share data. Even so,data annotation still represents a bottleneck, particularly in medicine andsurgery where clinical expertise is often required. With these constraints inmind, we propose FedCy, a federated semi-supervised learning (FSSL) method thatcombines FL and self-supervised learning to exploit a decentralized dataset ofboth labeled and unlabeled videos, thereby improving performance on the task ofsurgical phase recognition. By leveraging temporal patterns in the labeleddata, FedCy helps guide unsupervised training on unlabeled data towardslearning task-specific features for phase recognition. We demonstratesignificant performance gains over state-of-the-art FSSL methods on the task ofautomatic recognition of surgical phases using a newly collectedmulti-institutional dataset of laparoscopic cholecystectomy videos.Furthermore, we demonstrate that our approach also learns more generalizablefeatures when tested on data from an unseen domain.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答