Abstracts

Rationale: Temporal lobe epilepsy patients (TLE) with hippocampal sclerosis (HS) have reduced hippocampal volume in the diseased hemisphere. The atrophic mesial temporal structures produce excessive excitatory outputs that result in seizures. To understand how white matter connectivity relates to the disparity between volume and neural output, we used probabilistic and deterministic diffusion tensor tractography to quantify connectivity of the sclerotic and non-sclerotic hippocampus of TLE patients. Methods: Ten TLE patients (2 left, 8 right hemisphere HS) were imaged at 3 tesla (Philips Intera). A T1-weighted MRI and diffusion images (32 directions) were obtained. Automated anatomical parcellation (Freesurfer v4.5.0) of the T1 MRI was used to segment the diseased hippocampus (Dhipp) and non-diseased hippocampus (NDhipp) into separate binary image masks, which were used as seed regions for tract generation using both deterministic (DTI Query v1.1) and probabilistic (FSL v4.1.4) methods. For each patient, an index of inter-hemispheric hippocampal volume difference was computed as: [(NDhipp_vol-Dhipp_vol)/(NDhipp_vol Dhipp_vol)]. Indices of inter-hemispheric hippocampal connectivity density differences based on numbers of tracts were computed as: [(NDhipp_ntracts-Dhipp_ntracts)/(NDhipp_ntracts Dhipp_ntracts)]. Indices of inter-hemispheric hippocampal connectivity strength differences based on mean fractional anisotropy (FA) for voxels representing the tracts were computed as: [(NDhipp_fa-Dhipp_fa)/(NDhipp_fa Dhipp_fa)]. Results: Hippocampus volume in the diseased hemisphere was on average 21% smaller (Dhipp_vol=3146 mm3 vs. NDhipp_vol=3972 mm3, p<0.008). A strong positive linear relationship was found between the inter-hemispheric hippocampal volume index and the connectivity density (# tracts) index (Figure 1, r=0.71, p=0.02 for deterministic and r=0.98, p<0.0001 for probabilistic). A negative relationship was found between the inter-hemispheric hippocampal volume index and the connectivity strength (mean FA) index (Figure 2, r=-0.72, p=0.02 for deterministic and r=-0.74, p=0.01 for probabilistic). The deterministic and probabilistic methods produced a pattern of very similar results (r=0.75, p=0.01 for density and r=0.83, p=0.003 for strength). Conclusions: Automated anatomical parcellation estimated reduced hippocampal volumes accompanying HS. Volume reduction corresponded to fewer white matter tracts connecting the sclerotic hippocampus to the rest of the brain. Importantly, however, the decrease in tract density did not correspond to a reduction in connectivity strength, as measured by mean FA in voxels representing the tracts. In fact, we found the opposite relationship, with mean FA tending to be higher in tracts connecting the sclerotic compared to the non-sclerotic hippocampus. Very similar results were obtained using deterministic and probabilistic methods. We conclude that morphological differences in hippocampus of TLE patients with HS correspond to distinct differences in white matter connectivity density and strength.