Diffusion Kurtosis Imaging in Temporal Lobe Epilepsy
Abstract
Epilepsy constitutes one of the most common neurological clinicopathological entities affecting approximately 1% of the general population. Temporal lobe epilepsy (TLE) represents by far the most common form of medically intractable focal epilepsy in adults. Surgical resection is the common form of treatment when lesions are clearly delineated, either from patient’s magnetic resonance imaging (MRI) structural scans or by invasive seizure monitoring techniques (e.g., intracranial EEG) for patients with non-lesional MRI scans. Increasing numbers of studies have suggested that TLE is more of a network disorder, therefore full delineation of pathological tissue is difficult resulting in incomplete resection, possibly contributing to long-term recurrence of seizures after surgery. Diffusion MRI, an advanced MRI technique that is sensitive to the tissue at the microstructural level, has been studied, hoping to detect subtle microstructural changes related to TLE.
In this thesis, we investigated the ability of a diffusion MRI model, called diffusion kurtosis imaging, (DKI) to quantify TLE patients brain microstructure. Each chapter discusses the method developed to accomplish this, beginning with Chapter 1 giving the general background and the motivation behind this thesis. Chapter 2 develops a method of assessing the reproducibility in whole-brain high-resolution DKI at varying b-values. A shorter protocol was identified with comparable precision as the protocol with three b-values, supporting DKI for aiding clinical tools to assess brain tissue microstructure. Chapter 3 focuses on identifying microstructural abnormalities in the white matter (WM) and grey matter (GM) of the temporal pole, a region underappreciated in TLE patients. The method developed combining DKI measurements and tract-specific analysis uncovered temporal pole microstructural abnormalities in TLE patients (includes non-lesional TLE patients) compared to healthy controls. The work described in Chapter 4 explores a machine learning approach to laterialize TLE patients, demonstrating that DKI-based classifiers obtained slight increase in their general accuracy for GM region. Finally, Chapter 5 discusses the contributions of the thesis and provide suggestions for future research.