Abstracts

Rationale:
Background: Drug-resistant epilepsy (DRE) affects nearly 50% of the 3-4 million Americans diagnosed with epilepsy. Responsive neurostimulation (RNS) is a promising treatment for DRE, but its utilization remains suboptimal due to challenges in interpreting complex, single-subject chronic intracranial EEG (icEEG) data characterized by circadian and multidien rhythms.




Objective: To develop and validate a novel data visualization approach for identifying patient-specific temporal patterns of epileptiform activity in icEEG recordings from RNS-treated DRE patients, and to assess its potential for guiding personalized treatment optimization. Our approach was previously validated using synthetic seizure data mimicking real seizure variability and rhythms, and its clinical utility was assessed through structured surveys with an epileptologist. In this study, we extend the application of our platform to real-world clinical patients to evaluate its performance and potential for guiding personalized treatment optimization.




Methods:
We applied our novel data visualization platform to chronic icEEG recordings from two RNS-treated DRE patients with bilateral hippocampal implants. The platform generates raster plots displaying the frequency of stimulation-triggering interictal activity, color-coded (normalized across max frequency in the time interval of interest) and scaled across the recording period, with days on the y-axis (1 plot per hemisphere per patient) and hours on the x-axis.




Results: The visualization platform revealed distinct patient-specific epileptiform activity patterns. Subject 1 exhibited symmetric, night-time predominant inter-ictal patterns (onset ~9 pm, continuing through ~7 am, Figure 1), while Subject 2 showed diffuse left hippocampal inter-ictal activity (~8-9 am through ~2 am) followed by a clear transition to right-predominant activity (~5-6 am) with superimposed multidien rhythms on ~10-20 day timescales (Figure 2). These insights guided patient-specific adjustments in medication timing and RNS programming. Ongoing follow-up will assess the impact on seizure frequency and RNS-recorded seizure activity.


Conclusions:
Our novel data visualization platform effectively identifies patient-specific temporal patterns of epileptiform activity in chronic icEEG recordings, providing clinically relevant insights for personalized treatment optimization in RNS-treated DRE patients. Follow-up with patients will be maintained to assess subjective reports of changes in seizure frequency and to compare these with any concomitant decreases in recorded seizure activity on each subjects RNS. If successful, further validation in larger cohorts and assessment of long-term clinical outcomes are warranted.




Significance: By leveraging data visualization approach to improve the accessibility and interpretability of complex icEEG data, approaches similar to ours will have significant potential to enhance the utilization and efficacy of RNS therapy through patient-specific treatment optimizations. Beyond RNS, such platforms may also have wider applicability to other neuromodulation therapies and chronic EEG monitoring scenarios.




Funding: No funding