A Multi-View Mediator Framework for Subject-Specific Spatial Calibration in EEG: Stability, Sensitivity, and Cross-Condition Deviation

Abstract: Introduction: The spatial interpretation of scalp EEG remains constrained by volume conduction. In addition, the interdependence between spatial and temporal resolution is a barrier to EEG interpretation. Dominant approaches like source localisation, beamforming, and independent component analysis require structural MRI, manual judgement, or group-level templates. Further, analyses treat scalp location as absolute rather than relationally structured. The limitations of current methods and their dependence on costly imaging, manual interpretation, and reliance on group-level templates are a barrier to individual variation. A clear research gap exists in the provision of robust, interpretable, and personalised EEG spatial calibration. To address this gap, I propose a subject-specific, data-driven calibration framework that derives spatial information directly from each individual's EEG data, enabling more accurate, clinically relevant spatial interpretation without requiring structural MRI or group templates. Methods: Reconstructed topography is interrogated through three complementary views. The field, the inverse field, and the transport field are unified into a single channel selection (spTRIO) by a cross-exclusion mediator. State change is quantified through the following orthogonal deviation metrics: Basin shift, extremum displacement, and Jaccard distance. Taken together, all measurements can distinguish genuine neural reorganisation from artefacts. The pipeline was validated on the PhysioNet EEGMMI corpus, with all parameters held constant. Results: Spatial and temporal fidelity exceeded r = +0.94 across four frequency bands. (α,γ,δ, and θ). In addition, cross-subject generalisation outperforms the baseline. A β-band ceiling was diagnosed, with evidence indicating that β is a centrally mediated coordination rhythm. Other bands show evidence of a localised amplitude phenomenon. This is strong evidence of planned coordination, and the model may require a coordination view. The framework offers an interpretable, MRI-free calibration suited to longitudinal and individual-level monitoring.