Multi-label Classification of Surgical Tools with Convolutional Neural Networks

TL;DR

This paper develops a convolutional neural network-based system to automatically detect 21 different surgical tools in cataract surgery videos, addressing challenges like class imbalance for real-time surgical assistance.

Contribution

It introduces a residual network architecture tailored for multi-label surgical tool detection and explores various design choices through extensive experiments.

Findings

Achieved effective multi-label tool detection in cataract videos

Addressed class imbalance issues in surgical image datasets

Provided insights into optimal network design for surgical tool recognition

Abstract

Automatic tool detection from surgical imagery has a multitude of useful applications, such as real-time computer assistance for the surgeon. Using the successful residual network architecture, a system that can distinguish 21 different tools in cataract surgery videos is created. The videos are provided as part of the 2017 CATARACTS challenge and pose difficulties found in many real-world datasets, for example a strong class imbalance. The construction of the detection system is guided by a wide array of experiments that explore different design decisions.