Back to search

University of British Columbia

A machine learning framework for temporal enhanced ultrasound guided prostate cancer diagnostics

Abstract

dc:description

The ultimate diagnosis of prostate cancer involves histopathology analysis of tissue samples obtained through prostate biopsy, guided by either transrectal ultrasound (TRUS), or fusion of TRUS with multi-parametric magnetic resonance imaging. Appropriate clinical management of prostate cancer requires accurate detection and assessment of the grade of the disease and its extent. Despite recent advancements in prostate cancer diagnosis, accurate characterization of aggressive lesions from indolent ones is an open problem and requires refinement. Temporal Enhanced Ultrasound (TeUS) has been proposed as a new paradigm for tissue characterization. TeUS involves analysis of a sequence of ultrasound radio frequency (RF) or Brightness (B)-mode data using a machine learning approach. The overarching objective of this dissertation is to improve the accuracy of detecting prostate cancer, specifically the aggressive forms of the disease and to develop a TeUS-augmented prostate biopsy system. Towards full-filling this objective, this dissertation makes the following contributions: 1) Several machine learning techniques are developed and evaluated to automatically analyze the spectral and temporal aspect of backscattered ultrasound signals from the prostate tissue, and to detect the presence of cancer; 2) a patient-specific biopsy targeting approach is proposed that displays near real-time cancer likelihood maps on B-mode ultrasound images augmenting their information; and 3) the latent representations of TeUS, as learned by the proposed machine learning models, are investigated to derive insights about tissue dependent features residing in TeUS and their physical interpretation. A data set consisting of biopsy targets in mp-MRI-TRUS fusion-biopsies with 255 biopsy cores from 157 subjects was used to generate and evaluate the proposed techniques. Clinical histopathology of the biopsy cores was used as the gold-standard. Results demonstrated that TeUS is effective in differentiating aggressive prostate from clinically less-significant disease and non-cancerous tissue. Evidence derived from simulation and latent-feature visualization showed that micro-vibrations of tissue microstructure, captured by low-frequency spectral features of TeUS, is a main source of tissue-specific information that can be used for detection of prostate cancer.

Degree

thesis:*
Name thesis:degree_name
Doctor of Philosophy - PhD
Level thesis:degree_level
doctoral
Discipline thesis:degree_discipline
Electrical and Computer Engineering
Grantor dc:publisher
University of British Columbia
Year dc:date
2018

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Azizi, Shekoofeh

Rights

dc:rights
Statement dc:rights
  • Attribution-NonCommercial-NoDerivatives 4.0 International
Language dc:language
eng

Identifiers

dc:identifier.*
Handle dc:identifier
http://hdl.handle.net/2429/66385

Chain of custody

source
Harvested from
University of British Columbia
Base URL
circle.library.ubc.ca/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
related terms
citation

Azizi, Shekoofeh. A machine learning framework for temporal enhanced ultrasound guided prostate cancer diagnostics. doctoral thesis, University of British Columbia, 2018. http://hdl.handle.net/2429/66385