Abstracts

Recently, analysis of fMRI timecourse has been utilized for epileptogenic tissue localization (Morgan et al., Neuroimage 2004 21(1) 473-81). However, the temporal and spectral characteristics of interictal fMRI timecourses are largely unknown. In this study, we applied a signal processing technique called cepstral analysis (Bogert et al., Proc. Symposium on Time Series Analysis, ch. 15, 209-43. New York: Wiley, 1963) on interictal fMRI data. A motive for using cepstral analysis is that it can potentially factor out the spike event timecourse from the hemodynamic response function (HRF)., Resting state fMRI data from intractable seizure patients (age 10 to 37) whose intracranial EEG results agreed with their structural MRI abnormalities were used for the analysis. For each voxel, cepstrum was computed by (1) computing the logarithm of the Fourier spectrum of the timecourse and (2) computing the inverse Fourier transform. We further computed the factored-out spectra corresponding to the HRF (or spike event timecourse) by (3) extracting the initial 15.5 seconds (or later) segment of the cepstrum, and (4) re-computing the Fourier spectrum. A spike activity index for each voxel was computed as (2M - L - H) / (2M + L + H) where L, M, H are average spike event spectrum amplitude at 0-0.01, 0.04-0.05, and 0.3-0.32 Hz respectively., The spike activity index correctly identified the suspected epileptogenic tissue (SET) in four (two temporal and two occipital lobe epilepsy patients) out of seven patients. In these four patients, the SET showed timecourses that resembled the canonical HRF. The figure shows the spike activity index map (red: positive blue: negative) and factored-out spectra (red: spike event spectrum, blue: HRF spectrum) at the SET for a temporal lobe epilepsy patient.
In the three patients for whom the positive peaks did not coincide with the SET (two temporal and one frontal lobe epilepsy patients), the SET resulted in highly negative spike activity index, and the timecourses showed relatively flat lines with large baseline shifts. Two of these patients had positive peaks in the midline. The other patient had a positive peak at the contralateral homologue of the SET., We conclude that spectrum of the fMRI timecourse contains information that helps epileptogenic tissue localization and cepstral analysis shows promise for localization.[figure1], (Supported by Epilepsy Foundation postdoctoral training grant.)