EEG response to hyperventilation as a biomarker of clinical response and longitudinal changes in brain functional connectivity induced by vagus nerve stimulation

Details

Supervisors

Absil, Pierre-Antoine ; El Tahry, Riëm

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil biomédical, à finalité spécialisée

Abstract

enObjectives Vagus nerve stimulation (VNS) is a neuromodulation therapy used as an adjunctive treatment for patients with drug-resistant epilepsy who are not suitable candidates for resective surgery. Despite its proven efficacy, the mechanisms underlying its therapeutic effects remain incompletely understood, and no reliable biomarkers are currently available to predict or assess patient response. Cortical desynchronization has been proposed as one of the main mechanisms of action of VNS and can be investigated through functional connectivity measures. Hyperventilation (HV) is a well-established activation procedure in routine EEG recordings, but its impact on functional connectivity remains unclear. In this study, we investigated the predictive value of hyperventilation-induced EEG synchronization compared with resting state and evaluated how VNS modulates these synchronization patterns longitudinally. Materials and methods Twelve adult patients with drug-resistant epilepsy were followed before and at one, three, and six months after VNS device implantation. High-density EEG recordings were acquired during resting state (eyes open and eyes closed) and hyperventilation, under VNS ON and OFF conditions. The global and regional weighted phase lag index (wPLI) was computed as a measure of phase synchronization and correlated with the VNS response using multiple assessment methods, including binary classification (> 50% vs < 50% seizure reduction), percentage of seizure reduction, and the recently developed Clinical-Research Response Scale (CRRS). Linear mixed-effects models were used to assess longitudinal and condition-specific effects. In addition, several regression models were trained using selected feature combinations to evaluate and compare the predictive value of each condition, with an emphasis on model stability. Results During hyperventilation, lower values of global and regional wPLI at the preimplantation visit in the beta band, and more specifically the lower part of the beta band (13-20 Hz), were associated with better response to VNS, particularly when assessed using the CRRS. Although significant resting-state findings were limited, the best-performing predictive model combined global delta-band resting-state wPLI with beta-band hyperventilation wPLI, yielding a mean absolute error of 1.887 on the CRRS scale ranging from 0 to 15. Longitudinally, a progressive decrease in delta-band wPLI was observed during the eyes-closed VNS OFF condition, which correlated with VNS response over time when assessed using the CRRS. During hyperventilation, longitudinal changes in wPLI were primarily observed in the theta, high-beta, and broadband frequency bands, with better clinical outcomes associated with increasing synchronization over time. No significant acute differences were observed between VNS ON and OFF conditions. Conclusion Hyperventilation reveals distinct EEG synchronization patterns associated with subsequent response to VNS. Preimplantation hyperventilation-induced wPLI in the beta band may serve as a predictive biomarker, providing complementary information to resting-state measures. The chronic effects of VNS on hyperventilation-induced synchronization remain less clear and warrant further investigation. Nonetheless, the observed longitudinal decrease in resting-state delta-band wPLI is consistent with a progressive neural desynchronization.