Improved Somatotopic Consistency of EEG Source Localization Using a Personalized Segmentation-Free Head Model.

Somatosensory evoked potentials (SEPs) are widely used to assess the responses of somatosensory pathways and the central nervous system; however, the spatial resolution of EEG-based source localization is limited. This study examined whether a personalized segmentation-free head model combined with a sparse inverse solver could resolve finger-wise somatotopy without relying on functional MRI or magnetoencephalography. Sixteen healthy volunteers received electrical stimulation of the median nerve, ulnar nerve, and individual fingers, while SEPs were recorded using a 65-channel EEG. For each subject, high-resolution MRI was used to construct both conventional tissue-segmented and segmentation-free head models. Forward fields were computed using a finite difference method, and P20/N20 sources were estimated using orthogonal matching pursuit. Localization differences were evaluated as distances from reference coordinates reported in fMRI studies and compared with those obtained using a standard FEM-based pipeline with a distributed inverse solution. The segmentation-free model consistently reduced localization differences relative to literature-based reference coordinates by up to 2.9 mm. Across all stimulation targets, source-location differences (10-15 mm) were comparable to or smaller than those obtained using the conventional pipeline. Finger stimulation revealed clear somatotopic progression from the thumb to the little finger. These findings indicate that personalized segmentation-free modeling improves the spatial consistency of EEG-based source localization across multiple somatosensory targets. Rather than implying intrinsic enhancement of EEG spatial resolution, the results highlight the contribution of forward-model refinement to reducing systematic localization bias and enabling noninvasive assessment of somatotopic organization in the primary somatosensory cortex.