Abstracts

Rationale: The important and unique goal of this work was to preoperatively predict the extent of influencing critical patches of epileptic tissue communicating via white matter during investigational responsive neurostimulation (RNS, NeuroPace, Inc). The ictal propagation pathways were determined for an individual subject enrolled in the RNS Pivotal Trial using subtracted postictal diffusion tensor imaging (spiDTI), a novel functional high resolution DTI protocol. The volume of cortical activation (VOCA) was then predicted presurgically by modeling the electric field generated in white matter for current delivered through bipolar cylindrical depth contacts. Such Modeling was based on the Poisson equation solved with the Fourier finite element method (FEM). An iterative process was used to position the VOCA model. Regional white matter conductances derived from DTI were used in the calculations. Postoperative validation of the presurgically modeled VOCA was determined by subtracted activated SPECT (SAS) measuring stimulation-induced transient blood flow changes. Methods: Preceding implantation of the RNS generator and bihemispheric depth leads in subject TS (age 20 yrs), two gapless SPGR MRI sequences of the brain were acquired and averaged. In addition, a postictal high resolution DTI (acquired 8 hrs after a stereotypic complex partial seizure without generalization secondarily) and an interictal DTI (where the last seizure occurred 4 days prior) were normalized, subtracted and co-registered to the subject's averaged SPGR MRI. Electrographically, stereotypic seizures were captured by video-EEG monitoring. Scalp electrode positions (n=75) were digitized using a Polhemus digitizer. Source analyses of the ictal onset were performed using the boundary element method (Curry v4.6). The source localization data were coregistered with the spiDTI. Using a PC workstation equipped with 64GB of RAM, depth lead geometry and its modeled electric field determined by FEM were rendered and positioned in a 3D mesh of the spiDTI coregistered to the SPGR-MRI (COMSOL Multiphysics). Approximately five months postimplant, validation of predicted neural tissue targets affected by stimulation were determined by SAS.Results: The preoperative lead placement plan was validated postoperatively. Specifically, the preimplant VOCA model for subject TS was compared with SAS-associated transient blood flow changes acquired 5 months postimplantation of the RNS generator and depth leads. These data were also compared with a preimplant subtracted ictal SPECT and concurrent source modeling of the ictal onset. Randomization of subject TS in the ongoing RNS Pivotal Trial currently prevents assessment of seizure control. Conclusions: This data set demonstrates the ability to generate an individualized preoperative electrode planning map to estimate ‘best-implant’ sites for RNS depth leads. Development of this work will lead to a better understanding of guiding treatment stimulation through white-matter pathways to potentially distant epileptic neural tissue with a minimum of electrodes and current.