Epstein, Sean Carlo;
(2023)
Computational experimental design for quantitative MRI.
Doctoral thesis (Ph.D), UCL (University College London).
Preview |
Text
Thesis final.pdf - Accepted Version Download (19MB) | Preview |
Abstract
This thesis presents contributions to the field of quantitative MRI (qMRI) computational experimental design (CED). qMRI experiments are constructed from experimental ‘building blocks’ (e.g. acquisition protocol, model selection, parameter estimation) which, when combined, map tissue properties to quantitative biomarkers. Each of these blocks presents experimental choices: which acquisition protocol, which model, which parameter estimation method. Together, these choices form experimental designs. CED is the in-silico process by which such designs are tailored to suit specific imaging applications. This work addresses three limitations with current CED practices. The first is that they are too narrow in their scope: they are unduly focused on acquisition protocol. qMRI is underpinned by model fitting, which relies on an appropriate choice of signal model and fitting method. This choice cannot be taken for granted: model fitting both depends on and influences the quality of the acquired data. This work argues that CED should not focus on acquisition protocol alone, but rather consider all experimental components in an end-to-end, holistic manner. The second limitation relates to the experimental evaluation metrics currently used in CED. Experiments are assessed on their ability to generate close-to-groundtruth biomarker estimates, rather than on these estimates’ ability to solve real-world tasks (e.g. tissue classification); there is a disconnect between evaluation and application. This work address this by proposing a CED method which assesses experiments on their task performance, and validates its assessments on two clinical datasets. The final limitation relates to the parameter estimation methods available to CED. Existing methods are task-agnostic; they cannot be tailored to the needs of a specific qMRI experiment. This work takes advantage of machine learning techniques to, for the first time, make this possible: by changing training labels, parameter estimation performance is shown to be adjusted in a task-specific manner.
| Type: | Thesis (Doctoral) |
|---|---|
| Qualification: | Ph.D |
| Title: | Computational experimental design for quantitative 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 Med Phys and Biomedical Eng |
| URI: | https://discovery.ucl.ac.uk/id/eprint/10176656 |
Archive Staff Only
 |
View Item |
