Decoding Human Episodic Memory Using Hippocampal and Cortical High Frequency Oscillations

Introduction: Discrete memory representations are stored in widespread cortical areas, and the encoding, consolidation, and retrieval of rich, episodic memories requires the coordinated activity of multiple cortical regions. High frequency oscillations (HFOs) are 60-200Hz fluctuations in local field potentials that are tightly correlated to neuronal firing and likely play a significant role in orchestrating the cortical synchrony that underlies memory function. The link between brain-wide spatial and temporal patterns of HFOs and human episodic memory has not been studied, including whether HFOs may be used to decode specific memories.

Methods: We recruited 32 patients with epilepsy undergoing invasive neuromonitoring for identification of seizure networks. Patients viewed a continuous 10-minute stimulus and then performed a free recall. We identified HFOs from non-pathologic hippocampal and cortical contacts. Patterns of HFOs during memory encoding were related to consolidation and retrieval processes.

Results: We detected HFOs in 197 hippocampal and 3257 cortical contacts. Presence and location of HFOs was found to fluctuate during stimulus viewing, with HFOs in primary sensory areas preceding HFOs in association cortices. We identified spatial and temporal patterns of HFOs and found that these patterns were specific for each viewed event (p < 0.001). Event-specific HFO patterns exhibited significant pattern reactivation during consolidation of that event (p < 0.001) and during successful memory retrieval (p < 0.01). Finally, events that exhibited a greater magnitude of pattern reactivation between event viewing and consolidation were more likely to be recalled (p < 0.05).

Conclusion : HFOs across the human hippocampal-cortical memory network assemble in spatial and temporal patterns that are initially shaped during memory encoding and re-emerge during memory consolidation and retrieval. These patterns show specificity for single events and can be decoded. Future work will define the granularity of HFO-derived memory decoding by evaluating the contribution of individual HFOs within patterns and by using implantation paradigms with increased electrode density.