Towards the Development of an Automated Assessment System for the Fundamentals of Laparoscopic Surgery Tests

Abstract

The objective of this paper is to provide an overview of projects carried out in the framework of a research collaboration between the Department of Electrical and Computer Engineering and the Department of Surgery, in automated performance assessment for laparoscopic surgery training and testing. This paper focuses on describing the development of deep learning algorithms for object detection and tracking along with computer vision algorithms for performance assessment of Fundamentals of Laparoscopic Surgery (FLS) tests. The proposed automated assessment systems are based on quantitative measurements and expert knowledge using fuzzy logic. The Intelligent Box-Trainer System (IBTS) was used to create videos of several FLS tasks with the assistance of the medical school's surgery residents. Deep Learning (DL) models were developed and trained for three main tests of FLS: Precision Cutting, Peg Transfer, and Suturing. We placed our deep learning models in a publicly accessible database over the internet. The precision of our results compares favorably with other published work and with more data extracted from new videos, the fuzzy logic-based assessment system can be fine-tuned for even better performance.