Year: 2021
Type: dissertation
Abstract: There has been significant interest recently in exploiting routinely collected clinical and medical imaging data to discover new prognostic markers in multiple cancer types. In this work, we explore the use of multi-modal machine learning (ML) for outcome prediction in head and neck cancer (HNC).We first develop a ML approach for survival prediction from multi-modal data in the presence of competing risks, based on an extension of the multi-task logistic regression framework. We then describe the results of a ML challenge to develop a prognostic model for overall survival using electronic medical records and pre-treatment imaging. Finally, we present a novel graph-based method to incorporate imaging characteristics of metastatic lymph nodes and their connectivity patterns in a prognostic ML model using Gated Graph Convolutional Networks. Our results show the potential of ML in combination with large, multi-modal datasets as a tool to guide personalized cancer care.
Source: TSpace (University of Toronto)
Author: Kazmierski, Michal
Institution: University of Toronto
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W6983290007
