Diffusion MRI Biomarkers of RNS Efficacy in Epilepsy
Abstract number :
2.464
Submission category :
5. Neuro Imaging / 5A. Structural Imaging
Year :
2025
Submission ID :
1376
Source :
www.aesnet.org
Presentation date :
12/7/2025 12:00:00 AM
Published date :
Authors :
Presenting Author: Rocelle Evangelista, MS – University of California, San Francisco
Ehsan Tadayon, MD – University of California, San Francisco
Carter Lankford, BS Expected 6/2026 – University of California, San Francisco
Saboo Krishnakant, PhD – University of California, San Francisco
Leo Sugrue, MD, PhD – University of California, San Francisco
Vikram Rao, MD – Department of Neurology and Weill Institute for Neurosciences, University of California, San Francisco
Ankit Khambhati, PhD – University of California, San Francisco
Rationale: Responsive neurostimulation (RNS) is a promising therapy for patients with medically refractory epilepsy (MRE), yet clinical outcomes remain highly variable and difficult to predict. Identifying preoperative biomarkers that forecast treatment response could help optimize patient selection and inform neuromodulation strategies. In this study, we investigated whether diffusion MRI features extracted from limbic white matter pathways and mesial temporal structures could predict seizure reduction in patients receiving hippocampal-targeted RNS.
Methods: We retrospectively analyzed 41 patients with preoperative diffusion and structural MRI who underwent hippocampal-targeted RNS implantation. Imaging features were derived from a combination of diffusion tensor imaging (DTI) metrics and volumetric measures extracted from subject-specific white matter tracts and mesial temporal structures. A pipeline incorporating atlas-based segmentation and tractography was used to compute fractional anisotropy (FA), axial diffusivity (AD), and regional brain volumes. Clinical response was quantified as percent seizure reduction, defined using a time-weighted average across all available follow-up visits (capped at 36 months post-implantation) to account for variability to available data across patients. ElasticNet regression was employed to identify imaging features associated with clinical response, with model performance assessed using leave-one-out cross-validation and bootstrapped confidence intervals.
Results: Statistically significant relationships were observed between seizure reduction and several diffusion and volumetric MRI features. Notably, lower FA in the left fornix (p = 0.008) and right fornix (p = 0.032) were associated with greater seizure reduction. In contrast, higher FA in the right uncinate fasciculus (p = 0.015) and greater right amygdala volume (p = 0.020) were positively associated with improved outcomes. The final ElasticNet regression model explained 46.5% of the variance in clinical response (R² = 0.4652; RMSE = 26.4)
Conclusions: Preoperative diffusion and structural MRI features show promise as predictive biomarkers of clinical response to hippocampal-targeted RNS in MRE. Statistically significant associations were observed between seizure reduction and microstructural properties of bilateral hippocampal efferents (fornix), limbic association pathways (uncinate fasciculus), and volumetric measures of the amygdala. These findings underscore the utility of personalized neuroimaging biomarkers for informing patient selection and optimizing neuromodulation strategies. Ongoing work will expand the study cohort to include recipients of neocortical-targeted RNS, model the role of large-scale network connectivity, estimate volume of tissue activated (VTA) using RNS parameters, and integrate multimodal imaging features to enhance predictive accuracy and clinical applicability.
Funding: This work is supported by the National Institute of Neurological Disorders and Stroke (grant no. R61 NS125568-01A1).
Neuro Imaging