Abstracts

Following observations that the blood oxygen level dependant (BOLD) changes associated with inter-ictal epileptic scalp EEG discharges could begin or even peak before the discharge itself, we conducted a study to examine this unexpected phenomenon. From a database of 143 combined EEG-fMRI scanning sessions from patients with epilepsy, we examined all data sets in which there was a strong positive activation (5 voxels with a t statistic above 3.1 and one above 4.8) and which had discharges with a uniform morphology throughout the two-hour scanning session. This resulted in 20 data sets, three of which were rejected for excess motion ([gt] 5mm), leaving 17 data sets from 16 patients (one was scanned twice). Data sets were divided into those in which some or all events were prolonged EEG discharges (bursts of one second or more; 9 data sets) and those with no bursts (events were all singe spikes/spike-wave; 8 data sets). Statistical t-maps were generated using seven hemodynamic response function (HRF) models, consisting of a single gamma function with a peak 1 or 3 seconds before the discharge, or 1, 3, 5, 7, or 9 seconds after the discharge, with a full width at half maximum of 5.2 s. For all regions of activation in each t-map, an HRF was calculated from the BOLD response. HRFs for data sets with isolated events were projected using two methods: a Fourier basis method and a deconvolution method. Data sets with bursts were only projected using the deconvolution method, since this separates the effect of the duration of the burst and allows the time to peak of the underlying impulse response to be calculated. Projected HRFs with a poor signal to noise ratio were excluded. The time to peak of the remaining HRFs was calculated. HRFs projected using the Fourier basis method were very similar in shape and timing to HRFs projected using the deconvolution method for the same activation. Five of eight data sets with isolated events and two of nine data sets with bursts had projected HRFs that peaked before the event, or up to one second after the event. These activations with early HRFs do not seem to be more or less correlated with the location of the discharges on the scalp than activations with later peaking HRFs. They also do not seem to occur in data sets with more events, in relation to the presence or location of lesions, the patients[apos] age at time of scanning, at age of epilepsy onset, or clinical syndrome. The early peaking HRFs demonstrate that the BOLD response may change prior to the onset of scalp discharges by a matter of seconds. Given the width of an HRF, a projected HRF that peaks at the time of the event or one second after has begun to rise 3 or 4 seconds prior to the scalp discharge. This suggests that significant neuronal and metabolic activity may be present seconds before even a brief inter-ictal scalp discharge. (Supported by Canadian Institutes of Health Research.)