A delay bound certified most early baseline 16–28 Hz SEEG score signs under a 1 ms model in 21 people

A constant relative channel delay can change a signed phase score used by a brain-computer interface. We give a curvature bound that covers every delay in a stated interval. We tested the method on all 35 released SEEG records from 21 people in OpenNeuro ds006519 version 1.0.0. The primary score was mean signed imaginary coherency at 16, 20, 24 and 28 Hz in an early one-second baseline window. Under a relative delay from -1 to +1 ms with fixed spectral magnitudes, 90.84 percent of scores kept their sign and 9.16 percent could reach zero. Records first received equal weights within each person. People then received equal weights. The descriptive 95 percent person bootstrap interval for the zero fraction was 8.84 to 9.49 percent. All 37,283 primary pair occurrences with absolute score at least 0.1 kept their sign. The mean lower bound retained 85.74 percent of mean nominal score size. All source files passed hash checks, all voltage values matched a fresh source rebuild, and summaries replayed exactly. The package contains the proof, code, voltage arrays, spectra, pair bounds, figures and a raw retrieval recipe. The real-arithmetic proof uses a fixed spectral delay model. The code uses ordinary floating point operations with a numerical guard. The study does not establish neural coupling, source clock faults or decoder accuracy.

stereoelectroencephalography · SEEG · imaginary coherency · phase score · channel delay · signal processing · brain-computer interface · timing sensitivity