Advisor(s)

Rachel Smith
Arie Nakhmani

Committee Member(s)

Arie Nakhmani
Benjamin Cox
Joon Kang
Rachel Smith
William Tyler

Document Type

Dissertation

Date of Award

6-1-2026

Degree Name

Doctor of Philosophy (PhD)

School

Joint Health Sciences (Interdisciplinary)

Department

Biomedical Engineering

Abstract

Surgical removal of the seizure onset zone (SOZ) can eliminate seizures in people with medication resistant epilepsy. However, successful surgical outcomes depend largely on accurately localizing the SOZ, which is challenging as there is no clinically accepted bi-omarker of the region. Currently, intracranial EEG monitoring, an invasive multi-day pro-cedure during which patients are implanted with EEG electrodes, is used to record pas-sively occurring seizures, and the recording is used to identify the epileptogenic network including the SOZ. However, only 50% of surgeries conducted following this invasive localization procedure result in seizure freedom for the patient. Thus, it is necessary to identify biomarkers of epileptogenicity that can reliably distinguish SOZ from non-SOZ regions. In this dissertation, we investigated two potential biomarkers of the SOZ: the cortico-cortical spectral response (CCSR) and neural resonance. We first calculated CCSRs from stimulation-evoked potentials and quantified the proportion of significant CCSR responses, significance fractions, across different time-frequency zones. We found that the highest significance fractions in the zone spanning the delayed time period (50-500ms after stimulation) and the high gamma frequency band (50-250Hz) correlated best to the SOZ, indicating that these features of the CCSR could be a biomarker of the SOZ. Then, because neural resonance is estimated from dynamical network models of neural response, we validated model-predicted resonance using the CCSR. Neural reso-nance can be represented as a frequency versus magnitude curve where peaks in the plot indicate “resonance.” Using frequency versus magnitude curves extracted from the CCSR, we found that model-predicted resonant frequencies matched dominantly ex-pressed frequencies in the CCSR, indicating that the models can reliably predict neural responses to stimulation. Lastly, we utilized neural resonance to guide patient-specific stimulation protocols to elicit seizures, as these seizures can help localize the SOZ. We demonstrated that stimulation at resonant frequencies calculated from the SOZ subnetwork may more relia-bly elicit seizures which could lead to the development of an accepted stimulation proto-col to elicit seizures to localize the SOZ. Together, this work contributes to identifying computational biomarkers of the SOZ that could increase the likelihood of post-surgical seizure freedom for people with medication resistant epilepsy.

Keywords

cortico-cortical spectral response;dynamical network modeling;neural resonance;seizure onset zone localization;single-pulse electrical stimulation;stimulation-induced seizures

Download

Available for download on Saturday, May 29, 2027

Share

COinS