Sine-adjusted voltage signs gave higher mean error than Pearson sources for separate 60-second 1–80 Hz bipolar SEEG Pearson maps in 36 patients

We tested correlation maps from two public interictal SEEG files per patient in ds004100 version 1.1.3. Each source map predicted the same Pearson target map in the other file. The primary test used 60 seconds, 1–80 Hz, disjoint bipolar channels, and pairs from different shafts. Mean squared error was 0.004555 for Pearson and 0.004891 for sine-adjusted median-centered voltage signs. The mean difference, Pearson error minus sign error, was -0.000336. The 95% patient bootstrap interval was [-0.000508, -0.000170]. The interval excludes zero. The sign method had lower error in 7 of 36 patients. All 72 raw prefixes passed hash checks. The package includes edge maps, packed primary signs, one raw fixture, code, provenance, and fixed sensitivity tests. The target is a measured map. The study does not test true neural coupling or clinical outcomes.

stereoelectroencephalography · SEEG · correlation maps · Pearson correlation · sign correlation · epilepsy · prediction error