Evoked and Spontaneous High-Frequency Biomarkers for DBS Surgery And Programming

WSSFN 2025 Interim Meeting. Abstract 0031

Resumen

Introduction: Beta oscillations are established biomarkers of rigidity and bradykinesia in Parkinson’s disease (PD).^1-3 However, beta-band exhibits considerable intra- and inter-patient variability and correlates only partially with the spectrum of PD motor signs or clinical response to DBS.⁴ More reproducible biomarkers, directly tied to network dynamics evoked by stimulation, are needed to optimize DBS frequency, waveform, and spatial patterns of stimulation for individual patients. Here we assess two promising biomarkers - high-frequency oscillations (HFOs)^5-7 and evoked resonant neural activity (ERNA),^8-12 examining their spatial distribution and co-occurrence in the BG.
Method: We recorded intraoperative LFPs from directional DBS leads in PD patients undergoing awake surgery. Passive and evoked potentials (EP), obtained after bursts of DBS-like stimulation, were recorded. Cohort 1, 29 subjects (University of Pittsburgh, 2017–2019); Cohort 2, 9 subjects (MGH, 2024–2025). We quantified resting-state power spectra and evoked responses, focusing on HFOs and ERNA in STN, GPi, and VIM. Spectral features were extracted via a custom implementation of specparam,^13,14 and EP were analyzed using MATLAB and R. Statistical analyses employed linear mixed-effects models. We developed two computational models of the BG network in Neuroblox: a biophysically detailed H-H model,¹⁵ and a Next-Generation Neural Mass Model.
Results: Both HFOs and ERNA were robustly detected in both the STN and GPi, but not in VIM. Within the STN, HFOs exhibited clear spatial clustering, with “fast” (300-350 Hz) oscillations localized ventrally to “slow” (250-300 Hz) oscillations. ERNA waveforms, measured across the ventral and dorsal ring contacts, were remarkably consistent across stimulation from directional segments, suggesting activation of a single circuit. Failure to evoke ERNA in the STN was predictive of suboptimal lead placement. Our computational models of the STN and GP network successfully recapitulated spontaneous HFOs and the temporal dynamics of ERNA.
Discussion: HFOs and ERNA coincide in frequency and co-occur across patients, suggesting a common neural substrate. We hypothesize that spontaneous HFOs reflect intrinsic STN<=>GP dynamics, which are enhanced by DBS, manifesting as ERNA. Our computational models capture these dynamics, providing mechanistic insight.
Conclusions: High-frequency intraoperative biomarkers could help confirm accurate lead placement. ERNA can be recorded in less than 200ms per trial, enabling time-efficient exploration of stimulation parameters. Model-based fitting of patient-specific HFO and ERNA data can enable in-silico optimization of DBS settings, streamlining DBS programming.