Novel evaluation of surgical activity recognition models using task-based efficiency metrics

• DOI: 10.48550/arxiv.1907.02060
• Publication Date: 2019-01-01
• Authors: Aneeq Zia; Liheng Guo; Linlin Zhou; Irfan Essa; Anthony Jarc
• Keywords: Jaccard index

Purpose: Surgical task-based metrics (rather than entire procedure metrics) can be used to improve surgeon training and, ultimately, patient care through focused interventions. Machine learning models automatically recognize individual tasks or activities are needed overcome the otherwise manual effort of video review. Traditionally, these have been evaluated using frame-level accuracy. Here, we propose evaluating surgical activity recognition by their effect on efficiency metrics. In this way, determine when achieved adequate performance for providing feedback via from tasks. Methods: We a new CNN-LSTM model, RP-Net-V2, 12 steps robotic-assisted radical prostatectomies (RARP). our model both in terms conventional methods (e.g. Jaccard Index, task boundary accuracy) as well novel ways, such accuracy computed instrument movements and system events. Results: Our proposed achieves Index 0.85 thereby outperforming previous prostatectomies. Additionally, show that identified RP-Net-V2 correlate with labeled clinical experts. Conclusions: demonstrate metrics-based evaluation is viable approach quantify efficiencies. believe results illustrate potential fully automated, post-operative reports.