Optimisation, validation and application of quantitative magnetic resonance image analysis techniques to study the endophenotypic potential of brain morphology in temporal lobe epilepsy

The motivation for this thesis was in investigating the heritability of brain structure changes that may be useful in simplifying complicated genetic studies in epilepsy. An essential step in realizing this goal is the establishment of reliable and robust tools forautomated MRI-based in-vivo segmentation and measurement of geometric properties of brain structures. A reliable automated technique is necessary for future large scale genetic studies, where large amounts of data are needed to be produced efficiently, where manual methods are simply not practical. For this reason, this thesis focuses on investigating the optimisation, validity and reliability of Quantitative magnetic resonance imaging (QMRI) techniques for the measurement of geometric properties of brain structures.

In this thesis it is proposed to use the mathematically unbiased manual technique of stereology to validate the brain structure volumes produced by an automated method. In this work, the impact of object shape on the efficiency of MRI-based stereological sampling is firstly investigated for volume estimation and recommendations are then made for the optimal sampling of 4 varying shaped brain structures; cerebral white matter, temporal lobe, hippocampus and entorhinal cortex.

Using these optimal parameters, stereological estimates are further examined to determine accuracy, repeatability and reliability of the manual volume estimates, before being compared with volumes obtained from the automated method, FreeSurfer. Results from this validation of the automated segmentation procedure demonstrated Freesurfer to be a reliable technique for hippocampal and entorhinal cortex volume measurements when compared with manual b a d stereology.

This automated method was then applied to investigate potential MRI-based brain structure in patients with clinically defined temporal lobe epilepsy (TLE). A number of sub cortical structural volumes and cortical based volumes and thicknesses were examined and results demonstrated a reduction of volume in cerebral white matter, hippocampus, thalamus, mesial temporal cortex and left entorhinal cortex. A preliminary study was then carried out to investigate the heritability and, hence, endophenotypic potential of these identified abnormalities in a cohort of TLE patients, their unaffected siblings and control subjects. Our findings suggest a significant decrease in cerebral white matter and left entorhinal cortex common to patients and their non-affected siblings and a significant relationship between sibling pairs of thalamus volume, suggesting that structural volume changes in TLE patients may contain a heritable factor.