Abstracts

Rationale: Memory dysfunction impacts 40-50% patients with epilepsy, affecting workplace participation, school performance, and everyday life. Interictal epileptiform discharges, or spikes, during verbal encoding have been associated with forgetting but their relationship to physiological memory processing is unclear. Epilepsy patients undergoing invasive electroencephalography (iEEG) for resective surgery provide an opportunity to dissect the fine-grained spatiotemporal dynamics of memory. We describe the time course of high gamma activity (HGA), a measure of local population firing, across the brain during encoding of cross-modal associations. Next, we examined the impact of spikes on successful memory encoding in regional and network level analyses. Methods: A computerized associative memory task (face + profession pairs) was used to probe memory function in epilepsy patients undergoing invasive EEG monitoring for surgical evaluation at NYU Langone Health. Electrode localization was performed using automated processes and expert review. An automated spike detection algorithm combined with expert review was used to identify spikes. Analysis focused on determining differences in the spectral-temporal features in the high-gamma activity (HGA) between successful and failed encoding as well as investigating the impact of spikes on successful encoding. Results: Eight patients undergoing invasive EEG were included. Both successful and failed encoding trials demonstrated an early peak (<0.5 s) and sustained (0.5-2s) HGA in primary visual (occipital), visual association (inferior parietal), and high-level visual recognition (fusiform) cortex. Successful trials demonstrated greater HGA during late (1-2 s) encoding in bilateral medial temporal memory regions (entorhinal cortex, hippocampus, and amygdala) compared to failed trials. Late (1-2 s) differences in HGA in frontal executive and sensorimotor (primary motor, supplementary motor, pars triangularis/orbitalis, superior temporal gyrus) were associated with encoding performance. Right hemisphere demonstrated a different time course of HGA compared to left hemisphere, with greater differences in HGA distinguishing successful versus failed trials. Spikes occurring in bilateral medial temporal (L and R hippocampus, L entorhinal) during encoding were associated with a 2.0 greater odds of forgetting. Conclusions: In a cross-modal associative memory task, sustained neural activity in memory, executive/attentional, and sensorimotor networks characterized successful encoding. Differences in left and right hemisphere activity likely reflect specialization of visual processing of faces and words. Spikes in bilateral medial temporal lobe, which binds associations across modalities, disrupt encoding. Describing the physiological and pathological processes which support and disrupt associative memory may provide a therapeutic target for remediating memory function. Funding: NIH NINDS K23NS104252 (LIU)