Magnetic Resonance Imaging has been widely involved in detecting morphological brain changes in generalized epilepsy at voxel level. It is believed that region-based morphometry could provide better characterizations for macro structural changes in the brain with comparative advantages than voxel-based morphometry. The aim of this study is to detect regional brain changes associated with generalized epilepsy and to test the potential of regional brain changes in classification of patients with generalized epilepsy and non-epileptic subjects. 45 patients and 46 non-epileptic subjects were scanned using a 3 Tesla MRI scanner and 3D T1 weighted images were obtained. Images were pre-processed and region based structural metrics (grey matter volumes, white matter volumes, cerebrospinal fluid volumes, cortical gyrification, and sulcus depth) were developed using Computational Anatomical Toolbox (CAT12). Two sample t-tests were used to perform the univariate analyses and results were corrected for multiple comparisons (FDR corrected, p < 0.05). Grey matter volume reductions were detected in cerebellum, frontal lobe, temporal lobe, thalamus, and hippocampal region while bilateral white matter volume reductions were reported in cerebellum. In contrast, cerebrospinal fluid volume increments in left lateral ventricle were also detected. Reduced regional gyrification was detected in left posterior ramus of the lateral sulcus and reduced sulcal depths were detected in occipital pole, cuneus, and posterior ramus of the lateral sulcus in patients. Furthermore, pattern analysis revealed that each metric shows different discriminative abilities to distinguish patients with generalized epilepsy and non-epileptic subjects (Classification accuracy: 61.1%, 62.2%, 58.8%, 61.1% and 60% for GMV, WMV, CSFV, Cortical gyrification and sulcal depth respectively). In conclusion, this study provides a comprehensive understanding about regional structural changes (cortical and subcortical) associated with generalized epilepsy under region-based morphometry. However, pattern analyses do not provide adequate discriminative power and therefore the clinical utility of findings is limited.
Read full abstract