Computational Methods for Optimal Coil Placement and Maximization of Lobule-Focused Cortical Activation in Cerebellar TMS

ABSTRACT Background Coil placement on the cerebellum lacks accuracy in targeting the intended lobules and limits the efficacy of cerebellar transcranial magnetic stimulation (TMS) in treating movement disorders. Objective Develop a multiscale computational pipeline and method to rapidly predict the cellular response to cerebellar TMS and optimize the coil placement accordingly for lobule-specific activation. Methods The pipeline integrates 3T T1/T2-weighted MRI scans of the human cerebellum, lobule parcellation, and finite element models of the TMS-induced electric (E-) fields for figure-of-eight coils (MagStim D70) and double-cone coils (Deymed 120BFV). A constrained optimization method is developed to estimate the fiber bundles from cerebellar cortices to deep nuclei and, for both coil types, find the coil placement and orientation that maximize the E-field intensity in a user-selected lobule. Multicompartmental Purkinje cell models with realistic axon geometries and Gaussian process regression are added to predict the recruitment in the Purkinje layer. Results Our pipeline was tested in five individuals to target the left lobule VIII and resulted in normalized E-field intensities at the target 49.6±25.6% (D70) and 29.3±17.7% (120BFV) higher compared to standard coil positions (i.e., 3 cm left, 1 cm below the inion), mean±S.D. The minimum pulse intensity to recruit Purkinje cells on a 4 mm 2 -surface in the target decreased by 21.6% (range: 4.7-55.0%) and 10.7% (range: 7.9-18.2%), and the spillover to adjacent lobules decreased by 70.6±16.3% and 71.7±20.8% compared to standard positions (D70 and 120BFV, respectively). Conclusion Our tools are effective at targeting specific lobules and pave the way toward patient-specific setups.