Abstracts

Rationale: Neuropsychological evaluation of patients with Temporal Lobe Epilepsy (TLE) usually reveals a mild-to-moderate memory deficit. It is therefore important to build robust memory models in order to better understand cognitive deficits produced by TLE. The aim of this study was to build a model of memory encoding in healthy controls using fMRI data combined with independent component analysis (ICA) and Granger causality algorithm (GCA) so that this model can be applied to the evaluation of the effects of left and right TLE on the process of encoding. Methods: Forty healthy controls (39% female) aged 19-59 (mean age = 33) with no history of neurological disorders or memory complaints were recruited. All participants underwent fMRI at 4T in which they were administered a scene-encoding task. Participants were presented with stimuli that represented a balanced mixture of indoor (50%) and outdoor (50%) scenes that included images of inanimate objects and of people and faces. Participants were instructed to memorize all scenes for later memory testing. Control condition consisted of active comparison between two scrambled images. Compliance with the task was measured by asking the subjects to identify whether the presented scenes were in- or outdoor and by indicating whether the two scrambled images were identical. Within 10 minutes of completing the fMRI scan, participants were administered a post-scan recognition test that included 60 in-/outdoor scenes. Participants indicated whether they remembered seeing the picture in the scanner. Functional data were preprocessed using SPM8 (data were motion-corrected, normalized to an MNI template then spatially smoothed). Following preprocessing, twenty independent components (ICs) were computed from fMRI data by using Infomax algorithm (GIFT toolbox). Then, Granger causality algorithm (GCA) was applied to relevant ICs timecourses (timecourses were first detrended and demeaned) in order to reveal causal influence between them (GCCA toolbox).Results: Participants performed well on the scene-encoding task (M = 90.86, SD = 3.00 percent correct) and on the control condition (M = 93.25, SD = 4.10). After visual inspection, six task-related ICs were identified. These include IC1 (fusiform, parahippocampal, lingual, inf frontal gyri), IC2 (med frontal, ant cingulate), IC3 (sup/mid temp, inf/med frontal, ant cingulate and insula), IC4 (mid/sup temp, fusiform and parahippocampal gyri ad insula), IC5 (post cingulate, lingular gyrus and cuneus) and IC6 (occipital and posterior cingulate gyri). GCA revealed significant (p<0.05, FDR corrected) causal relations for all independent components except IC2. Conclusions: There is significant causality between components of the network for memory encoding indicating spatial and temporal dependence. These results provide a baseline model of memory encoding purely driven by fMRI data analysis (hypothesis-independent), which will help to better understand the effects of TLE on memory encoding.