Abstracts

Rationale: We recently developed the capability to perform simultaneous direct intracranial stimulation and fMRI, on patients undergoing invasive evaluation for intractable epilepsy. This technique can successfully produce maps of BOLD activation, both local and distal to the stimulating electrodes, whose patterns may reflect underlaying maps of connectivity. We now compare these connectivity maps to corresponding maps in the same patient: functional connectivity from electrode stimulation and recording, function connectivity from resting state fMRI, and structural connectivity derived from diffusion weighted MRI. Methods: To date, five patients with medically refractory epilepsy have been studied using the technique of direct electrical stimulation and simultaneous fMRI (DES-fMRI). All used a block paradigm for the temporal pattern of stimulation (typically 32 seconds at 8 mA/20Hz), with the recent addition of a successful event paradigm (2 second stimulation at 8 mA/20Hz). Connectivity measures from this method are simply the t-statistic of induced BOLD activation. EEG connectivity measures are derived from the mean recorded voltage obtained shortly after stimulation. After DES-fMRI, and prior to resective surgery, the patients underwent MRI for high resolution diffusion (61 direction) and resting state fMRI, with each providing a measure of connectivity. Thereafter, a total of four measures of connectivity are derived in the same patient, permitting a total of 6 paired comparison on a voxel-by-voxel basis. Pearson correlation values and significance were computed for each pair. Results: The first figure is an example of DES-fMRI connectivity derived using a 2 second stimulation as event paradigm (yellow arrow shows the location of the stimulating electrode pair, at 8 mA 20Hz). Each row shows a BOLD map at one second intervals, demonstrating the temporal evolution of BOLD signal changes across the brain. The second figure shows 6 scatter plots showing the paired correlation from the four different connectivity measures. The highest correlation is between the DES-fMRI t-statistic and EEG signal (R=0.20). All the remaining correlations show a smaller correlation coefficient, but all are statstically significant. Conclusions: We compare voxelwise connectivity maps made from four methods: DES-fMRI, EEG, diffusion MRI, and resting state MRI. While the correlations are statistically significant, they are modest. That is, there are many brain regions strongly connected by one measure, but weakly connected using another measure. These results raise questions about the meaning of connectivity, and in particular what is a "gold standard".