Abstracts

Rationale: Group analysis using voxel-based morphometry (VBM) have previously shown different patterns of grey matter (GM) atrophy in patients with non-lesional frontal lobe epilepsy (FLE), temporal lobe epilepsy with hippocampal sclerosis (TLE-HS), and temporal lobe epilepsy without hippocampal sclerosis (TLE-NL). This kind of analysis implicitly assumes these groups are homogeneous and that the abnormalities found at the group level correspond to those found in individual patients. However, this assumption is not necessarily correct. We therefore used cluster analysis to identify patients with typical and atypical GM atrophy patterns within each epilepsy group.Methods: 11 FLE (mean age 26.5 ± 7.3), 18 TLE-HS (mean age 41.8 ±10.2), 19 TLE-NL (mean age 40.5 ±10.1) and 23 healthy controls (mean age 35.1 ± 9.7) underwent 4T MRI, during which T1-weighted whole brain gradient echo images (TR/TE=3500/356ms, 1.0X1.0X1.0 mm3 resolution) were acquired. Diagnosis of epilepsy type was based on semiology and video-EEG monitoring; hippocampal subfield volumetry was used to determine presence / absence of HS on MRI. VBM (t-tests p<0.05 with age and intracranial volume as nuisance variables, no correction for multiple comparisons) was performed to compare patterns of grey matter loss between patient and control groups, and between patient groups, resulting in 9 different grey matter atrophy maps. Z score maps (<-2 cut-off) were also generated to describe atrophy pattern at the single subject level. Each patient’s Z score map was compared to each of the 9 group comparisons and Dice (similarity coefficient) scores were calculated. K-means clustering was used to reclassify all patients into 3 clusters, by matching the Dice scores representative of each patient’s atrophy pattern to the closest group prototype (FLE / TLE-HS / TLE-NL pattern). It was expected that for example a typical TLE-HS patient would have the highest dice scores in ‘Control>