Use this URL to cite or link to this record in EThOS:
Title: Penalised image reconstruction algorithms for efficient and consistent quantification in emission tomography
Author: Tsai, Yu-Jung
ISNI:       0000 0004 7660 4118
Awarding Body: UCL (University College London)
Current Institution: University College London (University of London)
Date of Award: 2019
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
With the increased interest in potential clinical applications based on quantitative results, the aim of this study is to improve the quantitative consistency of the reconstructed images in emission tomography (ET). To achieve practical processing time, a fast convergent quasi-Newton algorithm, preconditioned limited-memory Broyden-Fletcher-Goldfarb-Shanno with boundary constraints (L-BFGS-B-PC), is firstly proposed. Its performance is eval- uated using both simulations and three patient datasets. Results show that L-BFGS-B-PC is able to achieve several times faster convergence rate than separable paraboloidal surrogates (SPS). Moreover, the performance is less sensitive to penalty type, penalty strength, noise level and background level, compared to L-BFGS-B. To be able to improve the image quality and quantitative consistency, an anatomical penalty function is then considered with a spatially-variant penalty strength. Based on results for simulations and data from one patient with inserted pseudo lesions, the spatially-variant penalty reduces the quantitative dependence on the surrounding activity and location. Moreover, it benefits the algorithm convergence rate and its consistency among datasets. It is important to consider potential misalignment between the functional and anatomical images. For this reason, two approaches that perform alternating pe- nalised image reconstruction and misalignment estimation are therefore proposed. Expanding on the previous work, L-BFGS-B-PC using Parallel Level Sets (PLS) with the spatially-variant penalty strength is used in both approaches. Preliminary results for non-time-of-flight (non-TOF) data simulations demonstrate that both methods are able to estimate the misalignment and deform the anatomical image accordingly when a proper workflow for the alternating optimisation is applied. By integrating algorithms proposed in this study, both good image quality and consistent quantification can be achieved in a practical processing time.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available