RoHan: Robust Hand Detection in Operation Room

TL;DR

RoHan is a novel semi-supervised domain adaptation method that enhances hand detection in surgical environments by combining data augmentation with synthetic gloves and iterative refinement, reducing labeling needs.

Contribution

The paper introduces RoHan, a new approach that improves hand detection in operating rooms using semi-supervised domain adaptation and synthetic data augmentation.

Findings

Significantly improves detection accuracy in OR settings.

Reduces dependence on extensive manual annotations.

Effective across multiple surgical datasets.

Abstract

Hand-specific localization has garnered significant interest within the computer vision community. Although there are numerous datasets with hand annotations from various angles and settings, domain transfer techniques frequently struggle in surgical environments. This is mainly due to the limited availability of gloved hand instances and the unique challenges of operating rooms (ORs). Thus, hand-detection models tailored to OR settings require extensive training and expensive annotation processes. To overcome these challenges, we present "RoHan" - a novel approach for robust hand detection in the OR, leveraging advanced semi-supervised domain adaptation techniques to tackle the challenges of varying recording conditions, diverse glove colors, and occlusions common in surgical settings. Our methodology encompasses two main stages: (1) data augmentation strategy that utilizes "Artificial Gloves," a method for augmenting publicly available hand datasets with synthetic images of hands-wearing gloves; (2) semi-supervised domain adaptation pipeline that improves detection performance in real-world OR settings through iterative prediction refinement and efficient frame filtering. We evaluate our method using two datasets: simulated enterotomy repair and saphenous vein graft harvesting. "RoHan" substantially reduces the need for extensive labeling and model training, paving the way for the practical implementation of hand detection technologies in medical settings.