Abstracts

Rationale: We sought to determine if pathological ripple (pRipple) oscillations (80-200 Hz) can disrupt memory encoding. Methods: We detected and quantified ripple events in depth iEEG recordings in 117 DARPA RAM patients.  We used naïve Bayesian (NB) machine learning and a repeated measures linear regression model to quantify the relationship between pathological ripple events and the behavioral recall of a word following a distractor period. Results: pRipple events occurred very infrequently. The probability of ripple on epileptiform spike (RonS) events was increased in the seizure onset zone (SOZ, p < 0.05), and the probability of ripple on oscillation (RonO) events trended higher in the irritative zone (p=0.11). We found that all the types of pRipple events could disrupt encoding when they occurred at word presentation. Sharply contoured epileptiform discharges with significant ripple band energy, but without a true ripple oscillation, produced the strongest effect. The left and right temporal lobe, left parietal lobe, and right hippocampal formation proved the most susceptible (p < 0.05). In particular, encoding was most affected when pRipples occurred in the left middle temporal gyrus.    Conclusions: These results suggest that pRipples generated in certain regions can produce a dyscognitive state that interferes with memory encoding. Funding: This work was supported by the DARPA Restoring Active Memory (RAM) program (cooperative agreement N66001-14-2-4032).