(25) Computational Modeling of Human Vagus Nerve Stimulation with Three-Dimensional Fascicular Morphology

Introduction: Vagus nerve stimulation can treat several conditions—including epilepsy [1] and depression [2]—but its utility is constrained by side effects from activation of off-target nerve fibers [3], [4], [5], [6]. Computational models enable the design of selective stimulation approaches [7]. The traditional “extrusion” method extends a single 2D cross section to create a 3D nerve [8], [9], [10], [11], [12], overlooking the complex fascicular branching of the human cervical vagus nerve [13]. We created a computational pipeline to simulate fiber responses to electrical stimulation using 3D nerve morphologies.

Methods: We imaged human cervical vagus nerves using micro-computed tomography and segmented the nerve tissues (Figure 1, top). We used structural mechanics to model the nerve’s deformation within a cuff electrode and a fluid flow simulation to assign anisotropic conductivity. We defined 3D fiber pathways and simulated myelinated axon responses [14] to electrical stimulation from a cuff electrode. We compared the responses of these “true-3D” models to extrusion models [11] with cross sections from the same nerve samples (Figure 1, bottom). We systematically evaluated how model parameters affected the differences between paired extrusion and true-3D models.

Results: Extrusion models could accurately replicate true-3D model dose-response curves if the nerve was deformed, and the extrusion cross section was centered under the depolarizing electrode contact. Models using extrusion slices even a few millimeters from the electrode showed substantial discrepancies. We compared individual fiber thresholds and recruitment order between extrusion and true-3D models. For individual fibers, the median threshold percent difference was ≤12% across all fiber diameters. Using a multi-contact cuff electrode, both true-3D and extrusion models predicted strong spatial selectivity. The selectivity of extrusion models closely matched those of true-3D models when the selected cross section aligned with the active contact.

Conclusion: We addressed the longitudinal variability of the human cervical vagus nerve by creating a pipeline for true-3D modeling of peripheral nerve stimulation. Extrusion models replicated true-3D model vagus nerve stimulation responses if the nerve was deformed and if the extruded cross section was selected directly under the depolarizing contact. Small deviations from these conditions, however, caused large discrepancies in activation thresholds. Therefore, while extrusion modeling is an efficient and practical tool, fully volumetric approaches are advisable whenever electrode placement is uncertain, or multiple contacts deliver depolarizing stimuli. This work sets the stage for patient-specific modeling of peripheral nerve stimulation, enabling improved design strategies for vagus nerve stimulation that maximize therapeutic effects and minimize side effects.