Early Detection of Focal Seizures in Thalamus Using a Novel Seizure Detection Paradigm: Toward a Closed-Loop Thalamic Stimulation

Abstract

Despite therapeutic advances, many persons with epilepsy may not be candidates for resective surgery or RNS device if the seizure is non-localizable. One potential approach to treating epilepsy in this cohort is to provide stimulation to thalamus. One of the pre-requisite for developing an effective closed loop therapy is early detection of seizure. Here we report our experience in mapping the temporal dynamics of thalamic activity during transition from inter-ictal to seizures and compare a novel seizure detection method called "P-operator" to Curve length. Data was recorded from five adults with history of difficult-to-localize epilepsies using stereo-depth electrodes implanted in thalamus. Matlab was used to analyze power spectrum, power spectral density, comparison between P operator against curve length for substantiating functional connectivity between the seizure onset zone(s) and thalamus. A total of 22 seizures were recorded and selected for analysis. Matlab analysis with P-operator and curve length confirmed involvement of thalamus in 4 out of 5 patients. Curve length detected seizure earlier that P operator, however P-operator showed more efficacy for detection and count of interictals. Spectral analysis confirmed cortico-thalamic involvement between 100-400 Hz and the seizures were detected in thalamus within -10 to +5 seconds from cortical onset (T=0 sec). Thalamic signatures at seizure onset differed from offset. Two seizure detection methods (curve length and P-operator) gave similar results. Curve length had an average percentage of error of 5% compared to 9.6% for P operator. Thalamus was involved early in patients with intractable focal epilepsies.