Abstract:
EPILEPSY is a burdensome neurological disorder, which affects people of all ages. However, the modern age of science, medicine and technology, has substantially helped our understanding of epilepsy. Of many medical examinations, magnetic resonance imaging (MRI) is one most important non-invasive ways to analyze and identify structural and functional brain abnormalities. Along with the advent of mathematical/statistical based computational models, brain morphometry can be very well studied and understood.
Voxel-based morphometry (VBM) is a well-established computational approach, which uses MRI images to detect structural differences between groups of subjects or single subjects against normal controls, the latter commonly used to detect lesions in focal epilepsy. The aim of my research is to validate existing VBM methods and explore strategies to improve its performance in detecting subtle lesions in patients, which can be potentially epileptogenic. Improvements in existing VBM can aid clinicians in diagnosing abnormal brain morphology. However, a good sensitivity (related to true positives) and a reasonable specificity (related to false positives) is paramount for its clinical usage.
My work in epilepsy is categorized into three major studies, for the dissertation;
i. Validation-comparison of multispectral segmentation based on 3D T1-, T2- and T2-FLAIR and its performance in detecting visible cortical lesions (Lindig et al., 2017).
ii. Optimization of multispectral voxel-based morphometry models in identifying epileptogenic findings in MRI-negative patients (Kotikalapudi et al., 2018).
iii. Applications of MP2RAGE sequences for multispectral voxel-based morphometry for enhancing lesion detection in focal epilepsy patients with conventional MRI images (Kotikalapudi et al., in preparation).
In conclusion, computational analysis of multispectral (-contrast) MRI sequences via voxel-based morphometry, shows promise of improving lesion detection, especially in focal epilepsy patients with previously normal conventional MRI.
