Structural and effective brain connectivity in focal epilepsy

Epilepsy surgery is usually based on the removal of a local epileptogenic zone. If epilepsy is considered a network disease, a network approach might be more suitable. Insight into patient-specific epileptic brain networks is necessary to establish network-based surgical strategies. We included epilepsy surgery candidates who underwent diffusion-weighted imaging and intracranial EEG implantation with single pulse electrical stimulation (SPES, 0.2 Hz, 1-8 mA, 1 ms, monophasic stimuli) during presurgical evaluation. We reconstructed structural connectivity using fiber tractography taking intracranial electrodes as nodes. We reconstructed effective connectivity with SPES cortico-cortical evoked responses. We determined the inter-modal similarity between structural and effective connectivity with the Jaccard index, and compared network topologies using degree and betweenness centrality. We constructed a linear multilevel model to evaluate the relation between structural and effective connectivity at subject group level. The seizure onset zone nodes (SOZ), node proximity, and the volume of the electrode contact areas (VEA) were added to the model as possible predictors to accommodate for epilepsy and irregular spatial sampling. We included 13 patients (five with electrocorticography, eight with stereo-EEG). The median Jaccard index was 0.25 (IQR: 0.20-0.29), which means there is a higher overlap than expected by chance (median expected Jaccard index = 0.1 (IQR: 0.07-0.17)) with a considerable amount of connections that did not overlap. The structural connectivity degree showed a significant positive correlation with the effective connectivity degree in 9/13 patients and at group level after accommodating for node proximity (β = 0.13, 95 %-CI = [0.04, 0.21], t(852) = 2.79, p = 0.0054). SOZ and VEA were no significant predictors for the correlation between structural and effective connectivity. We showed a moderate overlap between non-invasive structural (measured with DWI) and invasive effective (measured with SPES) connectivity in epileptic brain networks. This overlap supports using non-invasively determined connectivity along with intracranial EEG to understand the epileptic brain. Future research needs to translate these findings towards network-based surgical strategies.