Electronic Thesis and Dissertation Repository

Thesis Format

Monograph

Degree

Master of Science

Program

Computer Science

Collaborative Specialization

Artificial Intelligence

Supervisor

Michael Bauer

Abstract

This thesis examines current state-of-the-art Explainable Artificial Intelligence (XAI) methodologies applicable to breast cancer diagnostics, as well as local model-agnostic XAI methodologies more broadly. It is well known that AI is underutilized in healthcare due to the fact that black box AI methods are largely uninterpretable. The potential for AI to positively affect health care outcomes is massive, and AI adoption by medical practitioners and the community at large will translate to more desirable patient outcomes. The development of XAI is crucial to furthering the integration of AI within healthcare, as it will allow medical practitioners and regulatory bodies to become more comfortable and trusting with respect to AI. The scope of this thesis is to examine XAI as it applies to breast cancer diagnostics specifically. However, as we have chosen to discuss local model-agnostic XAI methodologies, the techniques outlined in this thesis will be applicable to all medical domains.

Summary for Lay Audience

The main achievements of this thesis are as follows;

(1) Provide an in-depth technical overview of the theory behind state of the art local XAI methodologies

(2) Extensively apply local XAI methodologies to unveil the inner workings of a XGBoost black box model used to diagnose breast cancer with 96\% accuracy, using the Breast Cancer Wisconsin Diagnostic data set (BCW-D). This thesis is the most exhaustive analysis of local XAI methodologies applied to breast cancer diagnostics to date.

(3) Present a novel modification of the Biased Kernel SHAP algorithm called Fixed Biased Kernel SHAP, used to efficiently and accurately approximate true Kernel SHAP values, and evaluate the performance of this algorithm as compared to the original Biased Kernel SHAP algorithm.

Download

Share

COinS