Three calibration pulses gave 77.7 percent mean patient coverage for later 20 to 500 ms SEEG responses at 86 selected sites in eight patients

We tested limits for the full recorded response after brain stimulation. Public SEEG data supplied 86 selected sites in eight epilepsy patients. Five earlier pulses selected each readout. Three separate later pulses fitted each set. The remaining 352 eligible pulses supplied the test. The window was 20 to 500 ms. A full rank set covered 77.7 percent with equal patient weight. A ball fitted to the calibration errors covered 39.1 percent. The paired gain was 38.6 percentage points, with a descriptive 95 percent patient interval of 34.7 to 43.2 points. The pooled full-set count was 265 of 352. The full set is larger by construction. Its mean radial extent in fixed directions was 1.415 times the calibration-error radius. This is not a volume estimate. Before-pulse coverage was similar, at 77.1 percent. Acceptance therefore does not establish an evoked neural response. The 75 percent marginal rank guarantee requires exchangeability of the three calibration vectors and one future vector. The data do not establish that assumption. The ball geometry and rank guarantee are prior methods. This study adds a three-pulse benchmark after separate readout selection. The package preserves code, exact waveform data, source hashes, event roles, every test decision, fitted parameters, three figures, and reproduction instructions.

stereoelectroencephalography · SEEG · conformal prediction · brain-computer interfaces · electrical brain stimulation · waveform coverage · exchangeability · calibration