Chiou, Eleni;
(2023)
Domain Adaptation for Novel Imaging Modalities with Application to Prostate MRI.
Doctoral thesis (Ph.D), UCL (University College London).
Preview |
Text
PhD_Thesis_Chiou.pdf - Accepted Version Download (47MB) | Preview |
Abstract
The need for training data can impede the adoption of novel imaging modalities for deep learning-based medical image analysis. Domain adaptation can mitigate this problem by exploiting training samples from an existing, densely-annotated source domain within a novel, sparsely-annotated target domain, by bridging the differences between the two domains. In this thesis we present methods for adapting between diffusion-weighed (DW)-MRI data from multiparametric (mp)-MRI acquisitions and VERDICT (Vascular, Extracellular and Restricted Diffusion for Cytometry in Tumors) MRI, a richer DW-MRI technique involving an optimized acquisition protocol for cancer characterization. We also show that the proposed methods are general and their applicability extends beyond medical imaging. First, we propose a semi-supervised domain adaptation method for prostate lesion segmentation on VERDICT MRI. Our approach relies on stochastic generative modelling to translate across two heterogeneous domains at pixel-space and exploits the inherent uncertainty in the cross-domain mapping to generate multiple outputs conditioned on a single input. We further extend this approach to the unsupervised scenario where there is no labeled data for the target domain. We rely on stochastic generative modelling to translate across the two domains at pixel space and introduce two loss functions that promote semantic consistency. Finally we demonstrate that the proposed approaches extend beyond medical image analysis and focus on unsupervised domain adaptation for semantic segmentation of urban scenes. We show that relying on stochastic generative modelling allows us to train more accurate target networks and achieve state-of-the-art performance on two challenging semantic segmentation benchmarks.
| Type: | Thesis (Doctoral) |
|---|---|
| Qualification: | Ph.D |
| Title: | Domain Adaptation for Novel Imaging Modalities with Application to Prostate MRI |
| Open access status: | An open access version is available from UCL Discovery |
| Language: | English |
| Additional information: | Copyright © The Author 2023. Original content in this thesis is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0) Licence (https://creativecommons.org/licenses/by-nc/4.0/). Any third-party copyright material present remains the property of its respective owner(s) and is licensed under its existing terms. Access may initially be restricted at the author’s request. |
| UCL classification: | UCL UCL > Provost and Vice Provost Offices > UCL BEAMS UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Engineering Science > Dept of Computer Science |
| URI: | https://discovery-pp.ucl.ac.uk/id/eprint/10165221 |
Archive Staff Only
 |
View Item |
