Abstracts

Rationale: Analysis of the BOLD effect related to epileptic events is performed using the general linear model (GLM) with convolution between the event timings and the Hemodynamic Response Function (HRF). The model assumes that the baseline (time between each epileptic event) is stable during the entire recording. In practice, the vigilance state fluctuates during the fMRI acquisition and some patients fall asleep. This often alters the occurrence of epileptic events and we hypothesize that the variation of alertness, as reflected by the EEG, could interfere with the BOLD analysis.Methods: 1. BOLD effect of each rhythm. EEG-fMRI data (21 channels of continuous EEG inside the scanner) were selected from 11 patients with focal epilepsy having fewer than 6 spikes. The power of the EEG was calculated in the alpha band (averaged over posterior channels) and in the beta, delta, and theta bands (averaged over central channels). Six patients with spindles were also recorded and spindles were marked as events. All EEG analysis was performed after removal of the gradient and ballistocardiogram artefacts. The power fluctuations in each band and the spindle events were then convolved with the HRF and used as regressors in the GLM, resulting in a statistical t-map for each rhythm. A group analysis was also performed across subjects for each rhythm. 2. Effect of rhythms’regressors on the spike related BOLD signal. We recorded ten focal and ten generalized epileptic patients using the same method. Interictal spikes were used as events in the GLM analysis. We compared the t-maps obtained with and without using the rhythms’ regressors defined previously.Results: 1. There was no significant correlation between the BOLD signal and the beta and delta regressors. A positive correlation was found with alpha over the thalami, putamen and caudate nucleus. A strong negative correlation was also found over the bioccipital and bifrontal areas. For theta, a small negative correlation was observed over the frontal and parietal areas. The spindles were associated with a positive BOLD response over the thalami and the pallidum. 2. The statistical t-maps with or without the new regressors were not drastically different. We observed four types of results: a) improvement in 40% with an increase of the BOLD responses in locations coherent with the EEG signal. For the generalized patients, the default mode pattern was less apparent. b) decrease of the BOLD responses in 20% over locations not apparently related to the EEG field. c) no difference in 25%. d) mixed effects in 15%.Conclusions: Physiological modulations of the EEG rhythms during a period including wakefulness, drowsiness and sleep are clearly reflected in important BOLD signal variations in regions related to their generation. While considering these modulations in the analysis of spike related BOLD responses does not drastically alter the results, we often observed small improvements for focal and generalized discharges. This warrants the use of this new analysis, especially in patients showing great variations of physiological rhythms. (Rotary International, CIHR MOP-3879)