Separate thresholds at five checks raised background calls from 29.7% to 53.8% at 86 selected SEEG sites in eight patients with three calibration pulses

We tested repeated response checks after known stimulation pulses. The public source was OpenNeuro ds004696 version 1.0.1. We reused 86 selected contact differences in eight patients. Five pulses fit each reader. Three later background paths set its thresholds. The test had 353 later pulse pairs. Each patient had equal weight in the mean. One final check called 29.7% of background paths. Five checks with separate thresholds called 53.8%. One threshold for the full path maximum called 33.2%. Relative to separate thresholds, this rule reduced background calls by 20.6 percentage points and response calls by 3.2 points. It called 93.3% of response paths, including 84.4% by 100 ms. A rank proof gives a 25% marginal background bound for one future path under exchangeability. The data do not establish that assumption or demonstrate the bound. With three calibration paths, this deterministic rank rule cannot give a finite threshold at a 5% target. All eight processed source files passed full hash checks. All 1,042 saved pulse pairs matched a fresh extraction. The package contains input data, fitted readers, all scores and calls, three figures, the proof, and reproduction code. This is an offline measurement test in a reused clinical cohort. It does not measure continuous false commands or prove a neural origin for each response.

stereo-electroencephalography · SEEG · brain-computer interfaces · evoked responses · thresholds · rank calibration · background calls · signal processing · reproducibility