Offline response noise allocation had an uncertain 20 to 500 ms centered bipolar SEEG forecast gain at three pulses per site in eight recordings

We tested how to assign a fixed calibration pulse budget for stimulation response forecasts. The public SEEG data came from OpenNeuro ds004696 version 1.0.1. The test used eight recordings from eight people, 85 stimulation sites, and 879 site-to-readout pairs. Noise in the first two responses set the response rule. Noise before the same pulses set the control rule. Each rule used 255 early calibration pulses, with two to five per site. The final five pulses at each site gave 425 separate test targets. The mean reduction in normalized 20 to 500 ms waveform error was 3.30%. The descriptive 95% person-bootstrap interval was -0.42% to 6.94%. The two-sided sign-flip value was 0.141. Thus, the practical forecast gain remains uncertain. Early response noise was associated with later variance across sites. The mean rank correlation was 0.465 with the main scale and 0.380 without an energy divisor. The raw scale check was added after the planned results. Exact controls separate the count pattern from its assignment to sites. The package keeps compact trial products, four figures, pinned public-source hashes, validation results, and result tables. GPL analysis and extraction code is omitted, so the exact implementation cannot be rerun from this package. The data and test targets were used in prior studies. This is an exploratory offline reuse. It does not establish pulse savings or live brain-computer interface performance. Fresh rebuilds of all eight public source files matched every saved Gram matrix and source table. All 44 public metadata hashes also matched.

stereoelectroencephalography · electrical stimulation · response noise · pulse allocation · forecast error · brain-computer interface