Pooling chosen from five early pulses did not establish lower 20 to 500 ms SEEG map error than a shared map in seven sessions

We tested whether five early pulses could select useful pooling for later SEEG energy maps. We used public ds004696 version 1.0.1 and unchanged trial products from prior sci work. The test included 57 sites, 20 groups, and seven epilepsy sessions. The main window was 20 to 500 ms after a pulse. Each map gave the relative energy on the common remote recording shafts. The selected model had mean squared Hellinger error 0.018964; the shared model had 0.020568. The mean decrease was 0.001603, with a descriptive 95% patient interval from -0.000164 to 0.003363. The exact patient sign-flip reference gave p = 0.15625. The main test did not establish a gain or equal performance. A fixed half pooling model had lower observed mean error than the selected model. The package preserves maps, validation folds, pulse errors, three figures, source records, and reproduction code. Independent scalar checks and a full numerical replay passed. The test was retrospective within sites. It did not test online control, intended percepts, or safety.

stereoelectroencephalography · electrical stimulation · spatial energy maps · pooling · squared Hellinger error · brain-computer interfaces