Authors :
Presenting Author: David Greene, BS – NeuroPace, Inc.
Muhammad Furqan Afzal, PhD – NeuroPace, Inc.
Lise Johnson, PhD – NeuroPace, Inc.
Martha Morrell, MD – NeuroPace
Rationale:
Generalized tonic-clonic seizures (GTCs) are the primary outcome metric in the ongoing clinical trial to evaluate thalamic responsive neurostimulation in idiopathic generalized epilepsy (IGE, NAUTILUS trial, NCT05147571).
The RNS® System continuously monitors brain data, detects, and records instances of pre-defined epileptiform activity. The device also captures iEEG recordings associated with long duration detections (LEs), which are more likely to be electrographic seizures. Changes in LEs are predictive of clinical outcome in focal epilepsy, but in IGE patients with long trains of spike-and-wave discharges, the LE count may not reflect the clinical outcome of interest. In addition, these long spike-and-wave events may overwrite GTCs in device storage, making it difficult to capture the comparatively rare GTC recordings when they do occur. Therefore, detector settings with high selectivity for GTCs were developed and implemented in a subset of trial participants.
Methods: In the NAUTILUS trial, patients are implanted with bilateral leads targeted at the centromedian nucleus of the thalamus (CM). The GTC detector relies on simultaneous detection of low-frequency, high amplitude detection in both CM (bilateral high-amplitude detector, BHD). Device programming included a “Pattern A” detector to deliver therapy for the characteristic low-frequency delta activity prevalent in IGE that has been shown to correlate with absence seizures, and another much higher amplitude “Pattern B” detector that was generally specific for unequivocal electrographic seizures that correlated with GTCs (Figure 1).
Thirteen clinical trial patients were programmed with these detectors and kept a concurrent seizure diary. A concordance analysis was run to determine whether BHD detections corresponded to days on which GTCs were recorded in the seizure diary. A total of 228 BHD iEEG recordings from these patients were used for clustering to evaluate whether their electrographic patterns resemble typical GTC signatures. All four iEEG channels per recording for each patient were passed through a convolutional neural network to extract high-dimensional embeddings, which were subsequently reduced in dimensionality using PCA and t-SNE. Clustering was then performed using a Bayesian Gaussian Mixture Model.
Results: In three (3) of the participants, BHD detections were made, but GTCs were not recorded. In the remaining 10 patients, concordance was 100% for five (5), 90% for one (1), 75-79% for three (3), and 62% for one (1). Similarly, nine (9) out of the 13 patients showed iEEG channel clusters with characteristic low-frequency, high-amplitude activity consistent with GTC patterns (Figure 2).
Conclusions: This analysis demonstrates that a bilateral, high-amplitude detector implemented in CM may be an effective biomarker of GTCs in IGE patients.
Funding: Not applicable.