A Systematic Approach to the Development of an Automated Assessment System for Laparoscopic Surgery Fundamentals

This paper summarizes research projects conducted as part of a research collaboration between the Department of Electrical and Computer Engineering and the Department of Surgery aimed at automating the performance assessment of laparoscopic surgery trainees. A key objective of this paper is to describe the development of deep learning algorithms for the detection and tracking of objects, as well as computer vision algorithms that can be used for evaluating the performance of Fundamentals of Laparoscopic Surgery (FLS) tests. Based on quantitative measurements and expert knowledge, the proposed automated assessment systems use fuzzy logic to assess the effectiveness of the interventions. In collaboration with the surgical residents of the medical school, footage of several FLS tasks was created using the Intelligent Box-Trainer System (IBTS). A deep learning model (DL) was developed and trained to test three main aspects of FLS: precision cutting, peg transfer, and suturing. A publicly accessible database was created for our deep learning models. The results were compared based on the detection precision as different DL models were used. The mean average precision (mAP) for suturing task, peg transfer, and precision cutting was 97.6%, 85%, and 95.1%, respectively. These results have been demonstrated to be effective in detecting small surgical instruments used in laparoscopic box trainers, and for trainee evaluation purposes. Having more data extracted from new videos will allow us to fine-tune our fuzzy logic-based assessment system for better performance.