Abstracts

Rationale: Recent fMRI studies suggest that networks participating in cognitive funtion show correlated fluctuations of the BOLD signal at rest. Some patients with focal epilepsy show stereotyped onset zones and spreading network of seizure activity. We investigated if these pathological networks show correlated activity and a specific connectivity pattern in resting state fMRI recordings in patients undergoing epilepsy surgery for medically-intractable partial epilepsy. To test the feasibility of our methods to detect brain areas showing correlated BOLD fluctuations, we also computed connectivity maps for functional networks such as language. Functional and pathological networks were also defined using implanted subdural electrodes by recording ictal onset and spread patters as well as electrical stimulation mapping (ESM). Methods: Resting state fMRI was recorded on a 3T GE scanner in four patients with medically intractable focal epilepsy prior to implantation of subdural electrodes. Electrodes were localized using post-implantation CT and MRI followed by alignment to the surface of the pre-implantation MRI. BOLD signal was subjected to independent component analysis (ICA) as implemented in FSL (MELODIC) and functional connectivity analysis (RSFC). As seeds for RSFC we used region of interests below electrodes showing either epileptic activity and in two patients below electrodes that showed language function as defined with electrical stimulation mapping. Results: ICA components overlapping with subregions of the ictal onset zone were demonstrated in three patients. (an example is shown in figure 1). Similarly, only parts of the functional connectivity maps overlapped with epileptic networks defined with the subdural electrodes. The functional networks showed a trend to increased local connectivity and decreased connectivity with the contra-lateral hemisphere in two patients. RSFA reliably detected language areas showing the feasibility of our approach to detect functional resting networks. Conclusions: Parts of epileptic networks show correlated BOLD activity at rest. However, ICA and RSFC as used here showed little sensitivity to map the epileptic areas as defined with subdural electrodes. A combined approach, possibly including other methods such as single pulse cortically evoked potentials, that was recorded in these patients as well, might prove to increase the sensitivity. RSFC successfully localized anatomically and ESM defined language areas.