Abstracts

Rationale: The responsive neurostimulation system (RNS) for epilepsy can be programmed to record intracranial EEG at scheduled times throughout the day. These recordings capture neural activity that might be useful for understanding the electrophysiology of the targeted anatomical structures as well as RNS treatment effects. Spectral features of interictal recordings were estimated to characterize the electrophysiology of the subdivisions of the thalamus and hippocampus. Patient-reported outcomes were used to understand the relationship between anatomical location of interictal features and response to RNS. Additionally, modeling approaches were explored for understanding the relationship interictal neurophysiology and seizure activity over time.


Methods: Interictal recordings were obtained and identified using BRAINStim and iESPNet. The py_neuromodulation toolbox was used to estimate Fast Fourier Transform (FFT) and Sharpwave sharpness (non-sinusoidal waveform features) of recordings from patients implanted with RNS in the thalamus (n = 9) and bilaterally in the hippocampus (n = 6). Electrodes were reconstructed in MNI space and contacts were labeled using the THOMAS and Allen Brain atlas. Permutation tests were used to identify significant differences in feature means and standard deviations (STD) across anatomical label. The relationship between interictal features and responder status (defined by patient-reported change in quality of life) was examined for bitemporal patients. Features of interictal recordings from the baseline period were used to train a one-class Support Vector Machine (SVM) to evaluate whether interictal recordings were similar (i.e. an “inlier”) or different from baseline (i.e. outlier), and the relationship with outcomes was examined.


Results: There was a significant difference in the STD of Sharpwave features across labels (p < .025); STDs were higher for bitemporal contacts than thalamic contacts; there was no significant difference for FFT (see Figure 1, row 1: thalamic contacts = orange, bitemporal = blue). For bitemporal patients there was an interaction between responder status and label where the STD of Sharpwave features was significantly higher for non-responders than responders for contacts implanted in the body and head of hippocampus (p < .025) but not the white matter of the forebrain (see Figure 1, row 2: responder = green, non-responder = blue). There was no statistically significant relationship between the number of interictal recordings labeled “inlier” and “outlier” after the baseline period and outcomes.


Conclusions: Interictal recordings provide an opportunity for understanding the underlying electrophysiology of anatomical structures targeted by RNS. Non-sinusoidal waveform features provide unique information for understanding differences in physiology across different anatomical structures. Future work will continue to explore approaches for modeling the relationship between changes in neurophysiology and outcomes over time to understand whether interictal features can be used to characterize the impact of neurostimulation on pathological physiology.


Funding: National Institute of Neurological Disorders and Stroke, Grant # 5R61NS126776-02.