Abstracts

Rationale: Simultaneous EEG-fMRI has been investigated in epilepsy patients to localize BOLD responses to interictal epileptiform discharges (IEDs), by finding brain voxels whose fMRI signal is correlated with the IED timings convolved with a canonical hemodynamic response function (HRF). An underlying assumption is that IEDs with similar EEG morphology and localization are associated with the same HRF shape and amplitude. This study investigates more closely the relationship between the HRF and IED amplitudes/durations, which could help improve the model used in the analysis. Methods: EEG-fMRI data from 10 epileptic patients (40-75 min. at 3T, TR=1.75s, 5x5x5mm voxels) were selected on the basis of having a large number of clearly identifiable IEDs after gradient and ballistocardiogram artifact removal using averaged subtraction and independent component analysis (ICA), respectively. A neurologist selected, in an average reference montage, the channel where IEDs were most prominent and marked the start, peak, and end times of each IED. For spike-and-waves, the spike and the slow wave were marked separately. The markings were then fine-tuned by automatically adjusting them to the nearest EEG peak within 25ms. Subsequently, only IEDs separated by at least 4 seconds were considered to minimize overlap between closely spaced HRFs. The fMRI data was decomposed by ICA into spatially independent patterns with common time courses. A deconvolution method identified relevant components by detecting time courses with a significant consistent response time-locked to the IEDs. Individual variations from this deconvolved response were then calculated by fitting a gamma HRF to the time courses following each IED. Amplitudes of the fitted HRFs were correlated with IED amplitudes and durations (Spearman test, p<0.05). Results: 4 patients had only spikes, 2 patients had only sharp waves, and 4 patients had spike-and-waves. Except for 1 of the patients with spikes, the deconvolution of the ICA time courses could always identify one component significantly related to the IED timings and with spatial topography matching the presumed epileptogenic focus. In the other 3 patients with spikes, HRF amplitudes were not significantly correlated with spike amplitudes/durations. In the 2 patients with sharp waves, HRF amplitudes were significantly positively correlated with sharp wave amplitudes (fig.1). In the 4 patients with spike-and-waves, HRF amplitudes were significantly negatively correlated with spike amplitudes in 2 cases (fig.2), with slow wave amplitudes in 1 case, and with slow wave durations in the other case. Conclusions: The BOLD response to IEDs in the presumed epileptogenic focus is modulated by features that can be measured in the EEG recorded inside the scanner. This indicates that the neuronal events reflected by different IED morphologies result in measurable differences in metabolic responses. The analysis of EEG-fMRI data could benefit from integrating information about IEDs to improve the specificity of the model. HRF amplitude variations should thus be carefully investigated. Funding: NSERC CGSD, CIHR MOP38079