Abstracts

Rationale: Simultaneous scalp EEG-fMRI is a powerful tool that is quickly gaining popularity in both research and clinical settings. Combining intracranial EEG with fMRI (iEEG-fMRI) is of particular interest as it would allow interictal discharges to be recorded during fMRI with unprecedented precision. To this end, we have successfully implemented simultaneous iEEG-fMRI at 3T. In this study, we sought to determine the minimum number of epileptiform discharges required to generate consistent maps of BOLD activation during simultaneous iEEG-fMRI. Methods: Three iEEG-fMRI datasets were selected that had a high number of interictal discharges. All subjects were connected to a MR compatible EEG system, and simultaneous fMRI was performed at 3T using a GE Signa LX scanner. Functional images were collected using a series of single-shot GRE-EPI volumes, providing T2*-W contrast. In addition to the functional scans, the protocol included low resolution anatomical and T1-W high resolution anatomical scans. BOLD fMRI responses were modelled by convolving the timing of epileptiform events with a hemodynamic response function. Statistical maps were generated for the BOLD activation every 10 discharges until all recorded epileptiform activity had been modelled (i.e. 10 discharges, 20 discharges, etc.). The fMRI maps generated by all discharges recorded during the experiment were then used to contrast each discharge interval for significant differences (e.g., for Subject 1: 676 discharges vs. 10 discharges; 676 discharges vs. 20 discharges, etc.). Results: Subject 1 had independent bitemporal discharges that were modeled separately: 284 discharges from the left, and 194 from the right. We recorded 676 discharges from the temporal foci of Subject 2 and 820 from Subject 3's temporal foci. Upon analysis of the stratified data, it was found that the calculated locations of significant BOLD activation changed little after a certain threshold was reached for each subject: 96 discharges for Subject 1-left, 61 discharges for Subject 1-right, 90 discharges for Subject 2 and 180 for Subject 3 (Figure 1). As the number of interictal discharges included in the model increased, the statistical power of the BOLD activation increased and the changes became more localized with the subsequent elimination of noise. A contrast analysis was performed between the BOLD signal seen with the maximal number of discharges versus fewer discharges (Figure 2). There were few statistically significant differences and no large voxel cluster differences once 96 discharges were recorded for Subject 1-left, 61 for Subject 1-right, 50 for Subject 2 and 60 for Subject 3. Conclusions: Based on these data, a minimum of 50-100 interictal discharges need to be recorded during simultaneous iEEG-fMRI in order to produce reliable maps of BOLD activation. While as few as 10 discharges did result in localized BOLD activation, the activation was much more scattered and diffuse, with more noise. Although EEG-fMRI may produce maps of BOLD activation in subjects with relatively few discharges, these maps may not be accurate.