Use this URL to cite or link to this record in EThOS:
Title: Tissue characterisation in MRI : an assessment of the efficacy of textural features
Author: Boyce, David W. M.
ISNI:       0000 0004 2743 2882
Awarding Body: Imperial College London
Current Institution: Imperial College London
Date of Award: 1997
Availability of Full Text:
Access from EThOS:
Access from Institution:
This project considers the design and implementation of computational methods to assist clinicians in their interpretation of images from Magnetic Resonance Imaging (MRI). The ultimate aim of any diagnostic procedure is tissue characterisation. Conventional radiological image interpretation suffers from significant diagnostic inaccuracies. The research undertaken as part of this project specifically explores the use of sophisticated texture analytic methods to derive robust pathological indicators from MRI images with the ultimate aim of minimising diagnostic errors. A novel image visualisation and analysis environment is first developed utilising current graphics standards and user ergonomic features. This will form the basis of a clinically-oriented image processing platform for the evaluation and testing of the methods developed in this project. Three sets of texture analytic tools have been developed. The first set of algorithms is based on the image statistics and includes first order moments, run length, cluster size, grey tone difference and co-occurrence matrices. The second comprises of image transform methods and includes Fourier parameters and an original approach to the Walsh and Slant transforms. The third set is based on the measurement of the fractal dimension and its interpretation as a textural measure. The performance of the measures derived from each set of texture tools are evaluated when applied to sample sets of different texture classes from a library of clinical MRI images. The performance of the features is found from correlation and clustering measures as the image parameters are varied, for example; region of interest size, number of grey levels and data normalisation. Simple two feature tissue segmentation graphs are constructed to demonstrate the best performing texture measures. The results of this study enable the relative performances of the three sets of texture analytic tools to be compared.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available