Dual Invariance Self-training for Reliable Semi-supervised Surgical Phase Recognition

TL;DR

This paper introduces DIST, a semi-supervised learning framework that uses dual invariance principles to improve surgical phase recognition, effectively handling limited labeled data and reducing pseudo-label noise.

Contribution

The paper proposes a novel SSL method incorporating Temporal and Transformation Invariance, with a dynamic pseudo-label selection process for more reliable semi-supervised surgical phase recognition.

Findings

Outperforms state-of-the-art SSL methods on Cataract and Cholec80 datasets.

Enhances generalization to unseen data.

Robust pseudo-label selection reduces noise and improves accuracy.

Abstract

Accurate surgical phase recognition is crucial for advancing computer-assisted interventions, yet the scarcity of labeled data hinders training reliable deep learning models. Semi-supervised learning (SSL), particularly with pseudo-labeling, shows promise over fully supervised methods but often lacks reliable pseudo-label assessment mechanisms. To address this gap, we propose a novel SSL framework, Dual Invariance Self-Training (DIST), that incorporates both Temporal and Transformation Invariance to enhance surgical phase recognition. Our two-step self-training process dynamically selects reliable pseudo-labels, ensuring robust pseudo-supervision. Our approach mitigates the risk of noisy pseudo-labels, steering decision boundaries toward true data distribution and improving generalization to unseen data. Evaluations on Cataract and Cholec80 datasets show our method outperforms state-of-the-art SSL approaches, consistently surpassing both supervised and SSL baselines across various network architectures.