Toronto Metropolitan University
Browse
90fc29e23d44614268434f73b1682c49.pdf (4.67 MB)

Predicting Tumour Response With Radiomics and Machine Learning in MR-Guided Cervix Brachytherapy

Download (4.67 MB)
thesis
posted on 2024-03-18, 16:28 authored by Robert Bellis
This study seeks to determine if radiomic features extracted from whole or part of the gross tumour volume of locally advanced cervical cancer (LACC) patients can be used to predict tumour response prior to brachytherapy treatment. 12 machine learning algorithms were tested with 5-fold cross validation using 1183 radiomic features extracted from 20 patients from T1, T2 and diffusion-weighted MR images. Recursive Feature Elimination was used to indicate the most predictive radiomic features of the most accurate models. Several models, particularly Ensemble Methods, performed with accuracies of up to 85%. After combining the 11 most predictive features into a single dataset, a random forest model achieved an accuracy of 93%. Overall, this study showed that machine learning models coupled with radiomic features are capable of accurately predicting LACC tumour response prior to administering the first fraction of brachytherapy treatment.

History

Language

eng

Degree

  • Master of Science

Program

  • Biomedical Physics

Granting Institution

Ryerson University

LAC Thesis Type

  • Thesis

Thesis Advisor

Ananth Ravi and James Grafe

Year

2022

Usage metrics

    Biomedical Physics (Theses)

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC