Use this URL to cite or link to this record in EThOS:
Title: Quantitation in MRI : application to ageing and epilepsy
Author: Keihaninejad, Shiva
ISNI:       0000 0004 2700 4124
Awarding Body: Imperial College London
Current Institution: Imperial College London
Date of Award: 2011
Availability of Full Text:
Access from EThOS:
Access from Institution:
Multi-atlas propagation and label fusion techniques have recently been developed for segmenting the human brain into multiple anatomical regions. In this thesis, I investigate possible adaptations of these current state-of-the-art methods. The aim is to study ageing on the one hand, and on the other hand temporal lobe epilepsy as an example for a neurological disease. Overall effects are a confounding factor in such anatomical analyses. Intracranial volume (ICV) is often preferred to normalize for global effects as it allows to normalize for estimated maximum brain size and is hence independent of global brain volume loss, as seen in ageing and disease. I describe systematic differences in ICV measures obtained at 1.5T versus 3T, and present an automated method of measuring intracranial volume, Reverse MNI Brain Masking (RBM), based on tissue probability maps in MNI standard space. I show that this is comparable to manual measurements and robust against field strength differences. Correct and robust segmentation of target brains which show gross abnormalities, such as ventriculomegaly, is important for the study of ageing and disease. We achieved this with incorporating tissue classification information into the image registration process. The best results in elderly subjects, patients with TLE and healthy controls were achieved using a new approach using multi-atlas propagation with enhanced registration (MAPER). I then applied MAPER to the problem of automatically distinguishing patients with TLE with (TLE-HA) and without (TLE-N) hippocampal atrophy on MRI from controls, and determine the side of seizure onset. MAPER-derived structural volumes were used for a classification step consisting of selecting a set of discriminatory structures and applying support vector machine on the structural volumes as well as morphological similarity information such as volume difference obtained with spectral analysis. Acccuracies were 91-100 %, indicating that the method might be clinically useful. Finally, I used the methods developed in the previous chapters to investigate brain regional volume changes across the human lifespan in over 500 healthy subjects between 20 to 90 years of age, using data from three different scanners (2x 1.5T, 1x 3T), using the IXI database. We were able to confirm several known changes, indicating the veracity of the method. In addition, we describe the first multi-region, whole-brain database of normal ageing.
Supervisor: Hammers, Alexander ; Rueckert, Daniel Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral