Abstracts

Rationale: The direction and magnitude of interaction between brain regions can be quantified using functional MRI with Granger causality measures of temporal precedence. In this work we use these measures to determine influence between left and right mesial temporal regions in temporal lobe epilepsy (TLE) at rest. We hypothesize that these measures will reflect the order of seizure propagation proposed theories ictal networks of TLE, and may differentiate left TLE from right TLE. Methods: We acquired high temporal resolution fMRI images of TLE patients as part of a larger study involving lower temporal resolution fMRI. Images were acquired using a Philips Achieva 3T MRI scanner (Philips Healthcare, Inc., Best, Netherlands). The imaging series consisted of a T1-weighted structural MRI image, two low temporal resolution fMRI image sets (64x64, FOV = 240 mm, axial slices, TE = 35 ms, TR = 2 sec, 30 slices, 5 mm thick, 300 dynamics), and one high temporal resolution fMRI image set (same except TR = 500 ms, 9 slices, 400 dynamics). Conventional preprocessing and spatial normalization was performed. (1) Regions of interest (ROIs) - anatomic ROIs were determined using the left and right hippocampus regions of the Harvard Oxford probabilistic brain atlas. Functional ROIs were determined in the left and right mesial temporal lobes using 2dTCA [1] and functional connectivity [2]. These regions were part of a network of increased fMRI activity interictally in 4 left TLE subjects using low temporal resolution fMRI (TR=2s). (2) Region of interest identified in 500 ms TR data - The 500 ms TR data of 10 subjects (2 from functional ROI determination) lateralized by interictal and ictal scalp EEG/video monitoring (1 patient had normal EEG) was spatially normalized to the ROI templates. The average time course in each region was calculated for each subject. (3) Granger causality analysis - An order 1 autoregressive model was used to estimate model parameters for influence of previous time points of left ROI on current time point of right ROI (Influence LEFT), and vice versa (Influence RIGHT) for each patient. Motion parameters were used as confounds. The Influence LEFT-Influence RIGHT measures were then calculated. Results: We found a large variation of fMRI causality between the left and right TLE patients using the anatomic ROIs (Figure 1a), but a distinct separation between the two groups using the functional ROIs (Figure 1b). Presence of HS and PET hypometabolism was variable in these 10 subjects and not correlated with the fMRI causality measure. One patient in each TLE group was left handed. Conclusions: These results show that functional MRI with Granger causality can determine differences in cross hippocampal influence between left and right TLE patients if measured using functionally derived ROIs. Further interpretation and correlation to clinical measures are required. [1] Morgan, VL. Human Brian Mapping 2009;30:3393-3405. [2] Morgan, VL. Epilepsy Research 2010;88:168-178.