Abstracts

Rationale: Localization of the seizure onset zone (SOZ) to guide surgical treatment of medically refractory epilepsies is a challenge due to the lack of an accepted biomarker of this region. Intracranial EEG recordings of cortico-cortical evoked potentials (CCEPs) collected during single-pulse electrical stimulation (SPES) hold potential for SOZ localization as higher magnitude responses have been found in SOZ compared to non-SOZ regions. Further, time-frequency analysis of CCEPs to calculate cortico-cortical spectral responses (CCSRs) has shown that the strength of frequency expression of the evoked response following SPES is different between SOZ and non-SOZ regions. CCEPs can also be used to construct dynamical network models that estimate the relationship between the frequency of stimulation and the magnitude of the expected response. Neural resonance has shown potential for localizing the SOZ and guiding stimulation-induced seizures. In both CCSR and resonance analyses, there is a specific frequency associated with the maximum response of a brain region to a short pulse of stimulation. In this study, we test whether the resonant frequencies calculated from the models match the CCSR peak frequencies to validate the use of these methods to localize the SOZ.


Methods: All adjacent intracranial EEG contact pairs were stimulated during SPES (5mA, 1Hz, 30 trials) and responses were recorded at all other contacts. The data were re-referenced using a bipolar montage and stimulation artifacts and artifactual channels were removed. CCSRs were calculated by performing a continuous wavelet transform using a Morse wavelet (4-60 Hz), squaring the absolute value of the wavelet transform, and normalizing via decibel conversion. Resonant frequencies were calculated as frequencies at which a peak formed in the frequency versus magnitude plot (Bode plot) constructed from estimated dynamical network models. Response vectors from the CCSR coinciding with the N1 and N2 peaks were compared to the magnitude values from the Bode plot over the 4-60 Hz frequency range using the Pearson correlation.


Results: We expected that resonant frequencies calculated for an electrode contact would be expressed in the corresponding CCSR. In our preliminary comparison of 574 stimulation-response contact pairs from a representative patient, we found that the mean correlation between the N1 peak response and the Bode plot was 0.43 with a standard deviation of 0.39, and the mean correlation between the N2 peak response and the Bode plot was 0.69 with a standard deviation of 0.21 (Figure 1). We found that 23.5% of the N1 responses and 37.8% of N2 responses shared a peak frequency with a resonant frequency from the Bode plot.


Conclusions: Correlation of the peak response frequency of the CCSR and the resonant frequency validates the use of these dynamical network models to predict the brain’s response to stimulation. This finding reinforces the utility of neural resonance as a biomarker for the SOZ and as a guide for inducing seizures using stimulation.


Funding: This work was funded by the AES Junior Investigator Award 1042632.