Novel Computational Approaches for the Objective Analysis of Surgical Activities

Abstract

The primary goal of this dissertation is to advance the state-of-the-art in the objective analysis of surgical activities and to create novel metrics that can robustly and accurately compute the proficiency of surgical personnel. Toward these goals this work develops a range of computational approaches that takes advantage of a range of underused data available across surgical tasks both in an educational setting and in the operating room.Initially a novel approach to the analysis of surgical activities is presented that utilizes the inherent gesture-based structure of surgical activities and applies multiple kinematic-based metrics to the motion of surgical instruments. This novel granular approach is compared to traditional analyses over three different surgical tasks using regressions to evaluate the ability of the technique to classify subjective measures of proficiency. Overall this work presents significant evidence that a granular approach to the analysis of surgical activities is far better than traditional approaches.This dissertation also analyzes bimanual interaction, a component of surgical proficiency often discussed in literature that has yet to be addressed objectively in surgical activities. First, a general perspective is taken on the problem of evaluating the dependency between two or more systems by extending the non-linear systems concept of information transfer. Two novel extensions are presented in this chapter, first a multivariate extension that facilitates an evaluation of dependence between two or more multivariate systems, secondly a windowed extension is presented that facilitates the analysis of two or more multivariate systems whose dependency varies with time. These two extensions are evaluated on three unique simulated systems with results demonstrating their ability to accurately and robustly track dependence. Second, the non-linear systems concept of information transfer and the accompanying extensions presented in this dissertation is investigated as a metric in the analysis of proficiency in surgical activities. Multiple variants of information transfer are evaluated in both a traditional global formulation as well as in a granular analysis scheme like those discussed earlier in this dissertation. Regressions were used to evaluate the ability of information transfer in classifying subjective measures of proficiency. Results from these analyses showed strong evidence that such an approach is more discriminative than multiple existing state-of-the-art metrics and can offer unique insights in to the proficiency of surgical personnel.In summary, this dissertation presented multiple validated extensions to the state-of-the-art in the analysis of proficiency in surgical activities. The long-term goal of this work is to develop a closed system capable of robustly and accurately tracking intraoperative surgical proficiency and contributing to the education of surgeons by offering enhanced feedback, thereby optimizing the operating room and improving outcomes for patients.