Back to search

University of Sheffield

Automatic muscle segmentation from MR images using deformable registration and deep learning approaches

Abstract

dc:description.abstract

One in four people in the UK currently have one or more musculoskeletal disorder, impacting both work and social lives of individuals with them and incurring a burden on the national health service. Musculoskeletal disorders can affect one or both of the skeletal or the muscular system and where there is a substantial understanding of the mechanisms underpinning skeletal disorders, far less is currently understood regarding disorders affecting the muscular system. The challenge hindering our understanding of the mechanisms underpinning muscle disorders lies in the difficulty in measuring the physiological status of muscle tissue. Muscle disorders vary widely in many ways, such as the causes, muscles affected, rate of progression, and even the treatment strategies for these disorders. Not only do muscle disorders differ in these areas when comparing each one to the others, but also, people with specific muscle disorders respond to them in different ways. For these reasons, subject-specific, quantitative characterisation of the muscles within subject measured in vivo could enhance the current diagnosis and treatment strategies for muscle disorders. Moreover, quantitative tools to measure the response of the muscle tissue to new treatments for muscle disorders within clinical trials would grant a more informed analysis of the efficacy of treatments. Quantitative analysis of muscle tissue has not yet been adopted into clinical practice but could impact both our understanding and ability to treat people with muscle disorders. The aim of this thesis was to build, test, and analyse methods to automatically characterise the muscles from medical imaging data. Four methods have been detailed and explored to address the limitations associated with the current gold standard approach used to characterise the muscles from medical images. The outcome of the thesis is a general and complete overview of existing and novel methods to characterise muscles from medical images. Future work should analyse the methods presented in this thesis and adopt that which is best suited to their study. The motivation and ambition behind the thesis are that the studies presented facilitate future research seeking to understand muscle disorders in a quantitative manner. In the long-term, the work presented in this thesis could promote clinical adoption of computational tools for characterising muscle disorders, leading to enhanced diagnosis and monitoring of such disorders.

Degree

thesis:*
Name dc:type.qualificationname
Ph.D
Level dc:type.qualificationlevel
doctoral
Grantor dc:publisher.institution
University of Sheffield
Year dc:date.issued
2023

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Henson, William
Advisors dc:contributor.advisor
  • Mazzà, Claudia
  • Dall'Ara, Enrico
  • Xinshan, Li

Rights

Language dc:language.dcterms_iso639_2
eng

Identifiers

dc:identifier.*
Dc Identifier Grantnumber
EPSRCE MULTISIM GRANT - EP/K03877X/1, CM, EP/S032940/1, CM, ED

Chain of custody

source
Harvested from
White Rose University Consortium
Base URL
etheses.whiterose.ac.uk/cgi/oai2
Last updated
2026-11-18
Source record
OAI-PMH GetRecord
related terms
citation

Henson, William. Automatic muscle segmentation from MR images using deformable registration and deep learning approaches. doctoral thesis, University of Sheffield, 2023.