Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.764023
Title: Brain cortical variability, software, and clinical implications
Author: Mikhael, Shadia S.
ISNI:       0000 0004 7654 547X
Awarding Body: University of Edinburgh
Current Institution: University of Edinburgh
Date of Award: 2018
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Abstract:
It is essential to characterize and quantify naturally occurring morphometric changes in the human brain when investigating the onset or progression of neurodegenerative disorders. The aim of this thesis is to characterize the properties and measure the performance of several popular automated magnetic resonance image analysis tools dedicated to brain morphometry. The thesis begins with an overview of morphometric analysis methods, followed by a literature review focusing on cortical parcellation protocols. Our work identified unanimous protocol weaknesses across all packages in particular issues when addressing cortical variability. The next chapters present a ground truth dataset and a dedicated software to analyse manually parcellated data. The dataset (https://datashare.is.ed.ac.uk/handle/10283/2936) includes 10 healthy middle-aged subjects, whose metrics we used as reference against automated tools. To develop the ground truth dataset, we also present a manual parcellation protocol (https://datashare.is.ed.ac.uk/handle/10283/3148) providing step-by-step instructions for outlining three cortical gyri known to vary with ageing and dementia: the superior frontal gyrus, the cingulate gyrus and the supramarginal gyrus. The software, Masks2Metrics (https://datashare.is.ed.ac.uk/handle/10283/3018), was built in Matlab to calculate cortical thickness, white matter surface area, and grey matter volume from 3D binary masks. Characterizing these metrics allowed further understanding of the assumptions made by software when creating and measuring anatomical parcels. Next, we present results from processing the raw T1-weighted volumes in the latest versions of several automated image analysis tools-FreeSurfer (versions 5.1 and 6.0), BrainGyrusMapping, and BrainSuite (version 13a)- against our ground truth. Tool repeatability for the same system was confirmed as multiple runs yielded identical results. Compared to our ground truth, the closest results were generated by BrainGyrusMapping for volume metrics and by FreeSurfer 6.0 for thickness and surface area metrics. In conclusion, our work sheds light on the significance of clearly detailed parcellation protocols and accurate morphometric tools due to the implications that they both will have. We therefore recommend extra caution when selecting image analysis tools for a study, and the use of independent publicly available ground truth datasets and metrics tools to assist with the selection process.
Supervisor: Pernet, Cyril ; Hernandez, Maria Valdes Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.764023  DOI: Not available
Keywords: cortical ; variability ; morphometry ; grey matter thickness ; grey matter volume ; white matter surface area
Share: