A 50 percent calibration amplitude rule retained at least half the amplitude-only response size in four within-session SEEG tests

Can four early pulses select a response with a larger target share while limiting response loss? We tested a fixed rule in public OpenNeuro ds004457 version 1.0.2 SEEG data. The rule selected the largest target share among candidates with calibration root mean square voltage at least 50 percent of the largest target value. The first four eligible pulses selected a candidate. The last four separate pulses tested it. The test included 14 bipolar targets and 134 candidate choices in four patients, with one session per patient. All 14 later target responses stayed above half the amplitude-only response size. Selection by share alone put 8 of 14 below half size. The constrained rule retained a geometric mean size ratio of 0.907. Its equal-patient mean share gain was 0.0207 on a 0–1 scale, with a descriptive bootstrap interval from -0.0063 to 0.0532 and an exact sign-flip p value of 0.50. The same choices gave a before-pulse gain of 0.0258. Thus, the small share gain cannot be assigned to the stimulation response. The package supplies code, four figures, source-derived epochs, candidate mean waveforms, all scores, pulse allocations, and source retrieval hashes. All 110 raw files passed hash checks. Full raw extraction and a repeat analysis matched exactly. This offline test supports a separate response-size check when selection uses a ratio. It does not establish improved brain writing or online control.

stereo-electroencephalography · electrical stimulation · brain-computer interface · response selection · calibration · evoked response · response amplitude · OpenNeuro