New Transformer Framework Enhances Surgical Phase Recognition Accuracy

Researchers have developed a new framework for surgical phase recognition in laparoscopic cholecystectomy, integrating a Swin Transformer and Temporal Transformer Encoder (TTE) with a YOLOv11s network for instrument localization. This approach aims to provide stable and accurate workflow predictions, addressing the issue of fluctuating predictions in previous AI methods. The framework achieved 92% overall phase recognition accuracy and a macro average F1-score of 0.86 on the Cholec80 dataset, outperforming existing CNN and transformer-based methods. Additionally, the instrument detection module achieved an [email protected] of 0.863, demonstrating its potential for intelligent surgical assistance and robotic surgery applications. AI

IMPACT This research could lead to more reliable AI-powered surgical assistance, improving decision-making and documentation in operating rooms.

RANK_REASON The cluster describes a new research paper detailing a novel framework for surgical phase recognition. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• Cholec80
• Lövit
• Sheekar Banerjee
• Swin Transformer
• Temporal Transformer Encoder
• Trans-SVNet
• YOLOv11s

• paper
• model release
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →