Surgical Phase Recognition in Laparoscopic Cholecystectomy

TL;DR

This paper introduces a Transformer-based approach for recognizing surgical phases in laparoscopic cholecystectomy videos, using confidence scores to dynamically switch models, improving accuracy on the Cholec80 dataset.

Contribution

It presents a novel 2-stage inference pipeline that adaptively switches models based on confidence scores, enhancing surgical phase recognition performance.

Findings

Outperforms baseline model on Cholec80 dataset

Applicable to various action segmentation methods

Demonstrates improved accuracy in surgical workflow analysis

Abstract

Automatic recognition of surgical phases in surgical videos is a fundamental task in surgical workflow analysis. In this report, we propose a Transformer-based method that utilizes calibrated confidence scores for a 2-stage inference pipeline, which dynamically switches between a baseline model and a separately trained transition model depending on the calibrated confidence level. Our method outperforms the baseline model on the Cholec80 dataset, and can be applied to a variety of action segmentation methods.