Computational models for the treatment of refractory epilepsy with deep brain stimulation

Epilepsy is a chronic neurophysiological disorder that affects 1% of the world's population. While most of the patients are effectively treated with medication, surgical options can be explored, ranging from resective surgery to neuromodulation such as Deep Brain Stimulation (DBS), targeting the anterior nucleus of the thalamus (ANT). Despite its safety and effectiveness, ANT-DBS has a success rate of only 68%, posing several clinical and scientific challenges.
My research aims to improve the efficacy and the fundamental understanding of DBS in treating epilepsy by integrating personalized computational models at different spatial scales. We use a unique combination of clinical data from patients implanted with DBS, using scalp electroencephalogram (EEG) and simultaneous intracranial recordings (iEEG) from the thalamus.
We built personalized volume conduction models using scalp EEG data during DBS stimulation and performed source reconstructions on ANT-DBS evoked potentials. With simultaneous scalp and intracranial recordings (i.e., when the DBS is switched off), we designed a multiscale computational model based on individual scans of the patients. Using The Virtual Brain (TVB), we implemented two different models, a virtual DBS patient and a virtual epileptic patient. We fitted the relative parameters to the scalp and intracranial EEG, when the DBS was switched on or off, respectively.
Preliminary results consist of i) personalized DBS maps related to DBS evoked potentials for five patients and ii) replicated time-locked evoked activity induced by DBS on the scalp and the duration and amplitude of interictal epileptiform discharges (IEDs) recorded intracranially.
We will compare DBS maps to epilepsy maps resulting from reconstructed IEDs, and combine the virtual DBS and epileptic patients to explore the associations between the DBS settings and the efficacy of DBS therapy for epileptic patients.