Abstracts

Rationale: Data driven voxel based methods (VB) identify language networks variants in partial epilepsy patient subpopulations. Differences in activation occur in right homologues and adjacent to language processing areas but do not indicate the degree of language dominance. We compared VB with region of interest (ROI) methods. Methods: : 45 patients with a left hemisphere seizure focus (mean age 22.8, seizure onset 13.3) and 19 normal controls (mean age 24.8) performed a word definition 3T BOLD EPI fMRI language paradigm. Spatially normalized individual patient SPM maps were compared to the normal group in a voxel-wise comparison to identify voxels with z>2, followed by a principal component analysis, hierarchical clustering of variance patterns from individual difference maps, and k-means clustering into groups. ROI analysis of inferior frontal gyrus (IFG) and Wernicke’s Area (WA) using a bootstrap method was used to compute an asymmetry index (AI): Atypical language (AL) for each area was defined as AI<0.20. Results: We identified four patient groups from VB methods with the following mean AI for the ROIs: I) Increased left temporal activation on the margin of regions activated in controls (n=16) IFG 0.52 ± 0.37 (4 AL (25%)),WA 0.68±0.24(1 AL (5%)); II) No difference from normal controls (n=9), IFG 0.55±0.50 (2 AL (22%)), WA 0.69±0.34(1 AL( 12 %)); III) Recruitment in right IFG and temporal cortex (n=14), IFG 0.18±0.70 (7 AL, (50%),WA 0.41±0.67 (3 AL (22%)); IV) recruitment in right IFG and partially in temporal cortex (n=6), IFG -0.17±0.71 (5 AL (82%)), WA, 0.10±0.72 (3AL (50%)). Differences in AI among groups, except between groups 1 and 2 who were both left dominant, were significant (ANOVA, p<0.05). Conclusions: Using data driven methods we identify 4 different activation patterns that segregate with differences in measures of language lateralization. Two broad groups (left and atypical language) are identified, with the atypical language mostly driven by frontal rather than temporal language areas.