Abstracts

RATIONALE: With combined EEG-fMRI studies, it is possible to find brain regions activated as a result of epileptic spikes. A standard hemodynamic response function (HRF) is commonly used in fMRI activation studies (Glover 1999), but there are several reports showing variability in the shape of HRFs across subjects in sensorimotor activation studies. In addition, little is known about the HRF to epileptic spikes. The use of the patient-specific HRF may help identify areas activated by epileptic spikes. The objective of this study is to compare the activated areas obtained by using the Glover HRF and the patient-specific HRF.
METHODS: The HRF can be modeled as the sum of two gamma density functions (positive response and undershoot). We detected activated areas by comparing the fMRI signal at each point in the brain with a model based on the Glover HRF. For each activated area, we computed the actual HRF by averaging the BOLD signal around the time of the discharges. For these patient-specific HRFs, the positive responses are usually similar to the Glover HRF, but we found longer undershoot latencies and a variable ratio between the peak amplitudes of the positive response and the undershoot. We analyzed again the fMRI data using the patient-specific HRFs, and compared the activated areas and t-statistical scores to those obtained with the Glover HRF.
RESULTS: Six patients with intractable focal epilepsy were studied and all showed activation in the same areas with both methods, but the t-statistical scores were always higher with patient-specific HRF. The activated areas obtained with the patient-specific HRF were always wider. Also, additional activated areas were seen in three patients, two of them showing a significantly delayed undershoot. The first had bilateral subependymal heterotopia in the occipital horn and EEG spikes over the left temporo-occipital area. The Glover HRF showed activation in the left temporo-occipital junction only. With the patient[ssquote]s HRF, the activation included also the left subependymal heterotopia. The second had normal MRI and bifrontal spikes with clear right predominance. Activation was in the right frontal region with the Glover HRF but bifrontal with the patient[ssquote]s HRF. The third had right hemispheric atrophy maximum over the temporal region and bifrontal spikes. The Glover HRF showed several areas of activation in the left frontal lobe. An additional activation was seen over the right temporal lobe using the patient[ssquote]s HRF.
CONCLUSIONS: The patient-specific HRF yields wider activation areas and higher t-statistical values compared to Glover HRF in BOLD-fMRI studies of epileptiform discharges. Furthermore, new activation sites may be seen and these are concordant with what is known of the epileptogenic region in these patients. Using the patient-specific HRF brings new information in the analysis of EEG-fMRI studies of epileptic spikes.
[Supported by: This work was supported by the Canadian Institutes of Health Research under grant MOP-38079.]