Abstracts

Rationale: Epilepsy is increasingly considered a disease of brain networks, motivating the use of graph theoretical analysis to guide surgical planning in patients whose seizures are not adequately controlled by medication. However, incomplete sampling of epileptic networks due to sparse electrode placement may affect model results. Here we determine the sensitivity of several published network measures to electrode placement and propose an algorithm using network resampling to determine our confidence in the results of network analyses. Methods: We retrospectively analyzed intracranial EEG data from 28 patients who were implanted with a combination of grid, strip, and depth electrodes for epilepsy surgery planning. Upon both randomly and systematically resampling intracranial EEG electrodes, we recalculated global and local network metrics (Fig. 1).  Results: We found that sensitivity to incomplete sampling significantly varied between metrics (Friedman test: χ22 = 36.5, p < 0.001 for global metrics; χ24 = 107.9, p < 0.001 for nodal metrics), with transitivity and eigenvector centrality demonstrating the highest robustness to incomplete sampling amongst tested global and nodal statistics, respectively (Fig. 2). The sensitivity of network statistics to incomplete sampling was independent of the distance of the removed electrodes from the seizure onset zone. Finally, we demonstrate an algorithm using random resampling to obtain patient-specific confidence intervals on both global and nodal network theory statistics. Conclusions: The robustness of network statistics to incomplete sampling must be considered when performing network analyses using intracranial EEG data. Our findings highlight the difference in robustness between commonly used network metrics and provide tools that aid in the translation of personalized network models of epilepsy from bench to bedside. Funding: We acknowledge funding from NINDS R01-NS099348-01 (Litt; Bassett; Davis). Brian Litt also received support from NIH 1-T32-NS-091006-01 (Training Program in Neuroengineering and Medicine), The Mirowski Family Foundation, Neil and Barbara Smit, and Jonathan Rothberg. Kathryn Davis additionally received funding from NIH/NINDS (K23 NS073801) and the Thornton Foundation. Danielle Bassett would also like to acknowledge support from the Alfred P. Sloan Foundation, the John D. and Catherine T. MacArthur Foundation, and the ISI Foundation. Erin Conrad received support from NIH/NINDS R25-NS065745. Russell Shinohara received funding from the NIH (R01MH112845 and R01NS060910), the National Multiple Sclerosis Society, and the Race to Erase MS.