Use this URL to cite or link to this record in EThOS:
Title: Advanced magnetic resonance imaging and quantitative analysis approaches in patients with refractory focal epilepsy
Author: Kreilkamp, Barbara A. K.
ISNI:       0000 0004 7428 4640
Awarding Body: University of Liverpool
Current Institution: University of Liverpool
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Access from Institution:
Background Epilepsy has a high prevalence of 1%, which makes it the most common serious neurological disorder. The most difficult to treat type of epilepsy is temporal lobe epilepsy (TLE) with its most commonly associated lesion being hippocampal sclerosis (HS). About 30-50% of all patients undergoing resective surgery of epileptogenic tissue continue to have seizures postoperatively. Indication for this type of surgery is only given when lesions are clearly visible on magnetic resonance images (MRI). About 30% of all patients with focal epilepsy do not show an underlying structural lesion upon qualitative neuroradiological MRI assessment (MRI-negative). Objectives The work presented in this thesis uses MRI data to quantitatively investigate structural differences between brains of patients with focal epilepsy and healthy controls using automated imaging preprocessing and analysis methods. Methods All patients studied in this thesis had electrophysiological evidence of focal epilepsy, and underwent routine clinical MRI prior to participation in this study. There were two datasets and both included a cohort of age-matched controls: (i) Patients with TLE and associated HS who later underwent selective amygdalahippocampectomy (cohort 1) and (ii) MRI-negative patients with medically refractory focal epilepsy (cohort 2). The participants received high- resolution routine clinical MRI as well as additional sequences for gray and white matter (GM/WM) structural imaging. A neuroradiologist reviewed all images prior to analysis. Hippocampal subfield volume and automated tractography analysis was performed in patients with TLE and HS and related to post-surgical outcomes, while images of MRI- negative patients were analyzed using voxel-based morphometry (VBM) and manual/automated tractography. All studies were designed to detect quantitative differences between patients and controls, except for the hippocampal subfield analysis as control data was not available and comparisons were limited to patients with persistent postoperative seizures and those without. Results 1. Automated hippocampal subfield analysis (cohort 1): The high-resolution hippocampal subfield segmentation technique cannot establish a link between hippocampal subfield volume loss and post-surgical outcome. Ipsilateral and contralateral hippocampal subfield volumes did not correlate with clinical variables such as duration of epilepsy and age of onset of epilepsy. 2. Automated WM diffusivity analysis (cohort 1): Along-the-tract analysis showed that ipsilateral tracts of patients with right/left TLE and HS were more extensively affected than contralateral tracts and the affected regions within tracts could be specified. The extent of hippocampal atrophy (HA) was not related to (i) the diffusion alterations of temporal lobe tracts or (ii) clinical characteristics of patients, whereas diffusion alterations of ipsilateral temporal lobe tracts were significantly related to age at onset of epilepsy, duration of epilepsy and epilepsy burden. Patients without any postoperative seizure symptoms (excellent outcomes) had more ipsilaterally distributed WM tract diffusion alterations than patients with persistent postoperative seizures (poorer outcomes), who were affected bilaterally. 3. Automated epileptogenic lesion detection (cohort 2): Comparison of individual patients against the controls revealed that focal cortical dysplasia (FCD) can be detected automatically using statistical thresholds. All sites of dysplasia reported at the start of the study were detected using this technique. Two additional sites in two different patients, which had previously escaped neuroradiological assessment, could be identified. When taking these statistical results into account during re-assessment of the dedicated epilepsy research MRI, the expert neuroradiologist was able to confirm these as lesions. 4. Manual and automated WM diffusion tensor imaging (DTI) analysis (cohort 2): The analysis of consistency across approaches revealed a moderate to good agreement between extracted tract shape, morphology and space and a strong correlation between diffusion values extracted with both methods. While whole-tract DTI-metrics determined using Automated Fiber Quantification (AFQ) revealed correlations with clinical variables such as age of onset and duration of epilepsy, these correlations were not found using the manual technique. The manual approach revealed more differences than AFQ in group comparisons of whole-tract DTI-metrics. Along-the-tract analysis provided within AFQ gave a more detailed description of localized diffusivity changes along tracts, which correlated with clinical variables such as age of onset and epilepsy duration. Conclusions While hippocampal subfield volume loss in patients with TLE and HS was not related with any clinical variables or to post-surgical outcomes, WM tract diffusion alterations were more bilaterally distributed in patients with persistent postoperative seizures, compared to patients with excellent outcomes. This may indicate that HS as an initial precipitating injury is not affected by clinical features of the disorder and automated hippocampal subfield mapping based on MRI is not sufficient to stratify patients according to outcome. Presence of persisting seizures may depend on other pathological processes such as seizure propagation through WM tracts and WM integrity. Automated and time-efficient three-dimensional voxel-based analysis may complement conventional visual assessments in patients with MRI-negative focal epilepsy and help to identify FCDs escaping routine neuroradiological assessment. Furthermore, automated along-the-tract analysis may identify widespread abnormal diffusivity and correlations between WM integrity loss and clinical variables in patients with MRI-negative epilepsy. However, automated WM tract analysis may differ from results obtained with manual methods and therefore caution should be exercised when using automated techniques.
Supervisor: Keller, Simon S. Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral