Authors :
Presenting Author: Tal Benoliel, MD – Hadassah Medical Center
Atira Bick, PhD – Hadassah Medical Center
Avigail Waldman, MA – Hadassah Medical Center
Netta Levin, MD, PhD – Hadassah Medical Center
Dana Ekstein, MD, PhD – Hadassah Medical Center
Rationale:
Lateralization of memory functions in drug resistant epilepsy patients is used to predict post-surgical risk of memory impairment and may guide surgery. It is particularly important given well described variability in memory lateralization patterns in this patient population. fMRI paradigms have been gaining momentum but are limited in predicting verbal memory lateralization, often yielding insufficient activations and inconclusive results in memory related regions. Representational similarity analysis (RSA) is an fMRI analysis method aimed at identifying the sensitivity of a region to a certain condition, even when overall activation in the region does not differ significantly between conditions. Here we sought to apply RSA to a well-established hippocampal memory task, with the aim of improving the yield of the analysis.
Methods:
10 patients with drug resistant epilepsy underwent an fMRI task testing auditory and written verbal memory. Automatic parcellation was used to define regions of interest (ROI) in the hippocampus and limbic system and beta values were obtained. Beta value contrast between remembered and forgotten events and RSA were performed per patient, per ROI. A lateralization index was obtained from the RSA measurement by dividing the delta between right and left dissimilarity values by their sum. A value below 0.2 was considered bilateral.
Results:
Beta map contrasts between remembered and forgotten conditions were not higher in memory related regions. When applying RSA to memory related regions, significantly higher average dissimilarity values were found in the hippocampus, parahippocampal gyrus, entorhinal cortex and temporal pole compared with regions not associated with memory processing (calcarine sulcus, angular gyrus, frontal operculum) (p< 0.05). 5/10 patients had a memory lateralization index of above 0.2 in at least one hippocampal subregion, parahippocampal gyrus and entorhinal cortex for the auditory task, and 7/10 for the written words contrast, and were discordant in 2/10 patients. However, lateralization indices were not clearly unilateral in most tested ROIs.