A HAPwave prior reduced SEEG squared log ratio error by a mean 2.16% at 20 to 500 ms across reversed contact order with five fit pulses in eight STReEF patients

A prior from HAPwave reduced squared SEEG log ratio error by a mean 2.16% in eight selected STReEF patients. The ratio used voltage RMS at 20 to 500 ms after a pulse and -500 to -20 ms before it. The model used five local fit pulses. The five STReEF test pulses used the reverse contact order. All eight patients had lower error. A joint donor and target patient bootstrap interval was +1.19% to +3.36%. The reverse transfer raised mean patient error by 11.16%, with an interval that included no change. The top quarter of calibration ratios had no clear gain in either direction. A before-pulse control also gained in STReEF. Two pulses with the prior had 17.04% higher error than five local pulses in STReEF. The result supports a small offline calibration gain in this tested setting. It does not establish a stimulation-specific gain or a lower required pulse count. The package includes all selected measurements, forecasts, source records, code, a bias-cost proof, three figures, and reproduction instructions. An independent calculation passed. A separate replay reproduced all main result values exactly. The patients were used in earlier sci studies.

SEEG · electrical stimulation · voltage RMS · calibration · transfer learning · shrinkage · forecast error · epilepsy