Boundary Element and Finite Element Methods Show Consistent Results for Magnetospinography Forward Modelling

Magnetospinography (MSG) measures weak magnetic fields generated by spinal cord neural activity, offering a non-invasive window into spinal electrophysiology, but its accuracy depends heavily on the computational forward models used to interpret signals. This study tested four vertebral bone representations — continuous, homogeneous-toroidal, inhomogeneous-toroidal, and MRI-derived realistic — within both BEM and FEM frameworks. When comparing matched model pairs across the two methods, the researchers found strong agreement, with median relative errors below 3.1% and median squared correlation coefficients above 0.998, suggesting the choice of computational framework itself is not a major source of discrepancy. However, bone geometry proved consequential: segmented bone models (toroidal or MRI-derived) predicted substantially higher field amplitudes for left-right oriented sources compared to continuous bone models, with differences ranging from 35% to 72% across the cord. Notably, simplified toroidal models performed comparably to anatomically detailed MRI-derived models, with median r-squared values exceeding 0.97 between segmented model types, suggesting that full anatomical realism may not always be necessary. Sensor placement also influenced sensitivity, with posterior sensor arrays showing greater sensitivity in lower spinal regions, while anterior and posterior arrays performed similarly in the cervical region.