A loop check on 28 contact pairs missed gain faults in eight offline SEEG readers with no added output noise

We tested whether a loop check on SEEG contact pairs can detect errors that change a fixed task reader. We used public OpenNeuro ds004473 data from eight people and 1,128 later windows. We kept the published readers fixed. A +/-6 dB gain on one recorded contact gave no loop alarm in all 18,048 simulated cases. These faults changed 21.88% of decisions on average across people. They made 2,159 clean-correct decisions wrong. Matched errors on one pair output all gave an alarm. A proof shows that every error before pair subtraction is invisible to this check. The package contains a diagnostic tool, data, fixed models, scores, source records, and figures. The test had no added output noise. It does not measure real fault rates, online intent, or stimulation effects.

stereoelectroencephalography · brain-computer interface · signal integrity · fault detection · contact gain · pair differences · loop residual · offline decoder