Title:  Automated Benchmarking of Surgical Skills using Machine Learning

Committee: 

Dr. Essa, Advisor

Dr. Vela, Chair

Dr. Ploetz

Dr. Jarc

Abstract:

The objective of this proposed research is to design and develop an automated system for recognizing surgical activities and generating score reports in robot-assisted (RA) surgeries that can help surgeons move through their learning curves much faster. Currently, giving feedback to surgeons on RA surgeries requires manual segmentation of different tasks and evaluation of performance metrics. This approach can be extremely time consuming and tiresome especially since these procedures can be very long (> 1hr). In order to address this problem, I propose to (1) develop deep learning based models for recognizing different surgical tasks in a prostatectomy procedure using data from all streams coming from the da Vinci system, (2) evaluate meaningful performance metrics on segmented procedure to generate score reports, and (3) provide video highlights to surgeons from the procedure indicating which parts in specific effected their scores the most.