This study introduces a dataset of 140 surgical videos (MultiBypass140) of laparoscopic Roux-en-Y gastric bypass surgeries from two medical centers. The dataset has been fully annotated with phases and steps by two board-certified surgeons. The study assesses the generalizability of different deep learning models for the task of phase and step recognition. Results suggest that the model’s performance is influenced by the training data and that multi-centric training data improves the generalization capabilities of the models. The study concludes that multi-centric datasets are important for AI model generalization to account for variance in surgical technique and workflows.
Publication date: 19 Dec 2023
Project Page: https://github.com/CAMMA-public/MultiBypass140
Paper: https://arxiv.org/pdf/2312.11250
