Energy from different trials reduced contact rank agreement more before than after stimulation in seven SEEG patients with fixed source correction

We tested contact rank agreement in public SEEG data from OpenNeuro ds004696 version 1.0.1. Seven patients supplied 80 stimulation groups and 1,665 contact-group responses. Each group used its first ten good exported pulses. We tested all 126 partitions into two five-pulse sets. One measure used energy in the average waveform. The other used voltage products between different trials and kept negative values. Mean rank agreement before the pulse fell from 0.6885 to 0.0536. Agreement after the pulse fell from 0.8480 to 0.6795. The planned mean patient contrast was 0.4664. Its 95% t interval was 0.2953 to 0.6376. All seven contrasts were positive. The exact two-sided sign-flip p value was 0.015625. The test assumes independent patients and null sign symmetry. The result held in planned descriptive timing and contact checks. Under independent zero-mean trial noise and a common mean trace, the different-trial measure removes self-energy bias. This proof does not establish the noise assumptions in the source data. Source reference selection used all exported trials. The empirical result depends on that fixed correction. It does not identify neural connections or an adequate clinical pulse count. The package preserves code, compact data, voltage examples, source versions and hashes, results, and reproduction instructions. It uses the input cohort from AutoSci a8be6ea7-2028-46b1-b189-6d140f2b5d67.

stereoelectroencephalography · trial variability · contact rank agreement · electrical stimulation · energy estimator · cross-validation · signal analysis