Mean band maps agreed more between files than band differences after distance and node controls in 36 SEEG patients and 10 fMRI participants

We tested whether a frequency band contrast can predict a second recording after distance and recording-site effects are removed. We used separate public studies: 36 HUP SEEG patients and ten PAFIN fMRI participants. The fixed half-weight predictor reduced band-contrast error by 32.92% in SEEG (95% person bootstrap interval 26.54% to 38.47%). It increased error by 29.88% in fMRI (26.04% to 34.35%). Thus, the planned cross-modal transfer failed. A follow-up compared spatial agreement between files for the mean of two absolute correlation band maps and their difference. Mean-map agreement was higher in both studies. The mean paired cosine decrease was 0.1097 in SEEG (0.0861 to 0.1322; 35 of 36 people) and 0.2854 in fMRI (0.2303 to 0.3428; all ten people). The result held with non-overlapping SEEG signals, NORDIC fMRI, equal fMRI run lengths and squared correlations, with wider SEEG uncertainty. The follow-up was exploratory in fMRI and checked in a separate, previously studied SEEG cohort. It lacks new-participant confirmation in both modalities. Frequency bands, tasks, disease, age, spatial sampling and noise differ. HUP file overlap remains unknown. The result concerns measured map agreement, not a common neural frequency or causal network. Code, compact inputs, exact source records, all negative results, figures and repeat instructions are included.

human neuroscience · SEEG · fMRI · frequency bands · network reliability · cross-modal test · negative result · reproducible benchmark