Abstracts

RATIONALE: EEG recorded inside the MR scanner enables the analysis of fMRI data based on the timing of epileptic discharges. In order to detect the functional activation correlated with epileptic events, one needs to assume a model of the hemodynamic response (HR). Our objective was to compare the experimental HR with one commonly used model (Glover, Neuroimage 1999).
METHODS: We acquired seven to eight runs of 120 fMRI frames, each frame consisting of 25 64x64 echo planar images (5mm thickness, 3s repetition time) for two patients with focal epilepsy. All the measurements were performed on a 1.5 T Magnetom Vision whole body scanner. Simultaneously, we recorded 19 channels of EEG with a specially designed system (Schwarzer, Germany). Fourier filtering was applied to the EEG in order to reduce the fMRI-induced artefact (Hoffmann et al. Magn Res. Med. 2000). We marked epileptic events and performed statistical analysis with the HR model used for normal functional activation studies. The time course of the BOLD response was obtained on the regions with statistically significant activation. We applied third-order polynomial detrending to the signals and averaged them using epileptic events for temporal alignment.
RESULTS: Most epileptic discharges were followed by a clear BOLD response. The average HR[ssquote]s were similar to the model, although patient one presented no undershoot below the baseline and both were somewhat more extended in time than the model. The peak lags were around 6s and the return to baseline was approximately at 33s for patient one and at 24s for patient two. Patient two presented only bursts of spikes, and we did not observe a correlation between burst duration and HR duration or amplitude. We observed large very slow fluctuations in the non-detrended fMRI values inside runs and from one run to the next.
CONCLUSIONS: The hemodynamic response of epileptic spikes is quite similar to that observed in visual or auditory functional activation, although some small departures in shape or duration were observed. The concordance was expected, because the regions of activation are the ones that correlate well with the model. Nevertheless, we did not observe an increase in BOLD amplitude for longer spike bursts, as was expected. The important gradual fluctuations of the BOLD signal and the long HR could explain why former EEG-triggered fMRI experiments produced relatively few activations in our laboratory.
Support: Canadian Institutes of Health Research grant MOP-38079