An early gain check changed later joint SEEG agreement by 0.01 percentage points in eight sessions at a 50 percent tolerance

We tested an early gain check for public SEEG responses to single electrical pulses. The data came from eight epilepsy patient sessions in OpenNeuro ds004696, version 1.0.1. We used 886 eligible bipolar readouts. The first five good pulses selected readouts. The final five good pulses tested each selected readout. A match required correlation of at least 0.8 and signed gain from 0.5 to 1.5 in the 20 to 500 ms window. The shape check selected 125 readouts. Adding the gain check selected 114 readouts. Mean later joint agreement changed from 79.04 percent to 79.06 percent. Each patient had equal weight. The mean change was +0.01 percentage points. The 95 percent patient bootstrap interval was -1.27 to +1.00 points. The exact patient sign-flip p value was 1.0. These data showed no clear benefit from this added check. This result does not prove equal performance. The package includes pulse metrics, three figures, exact pulse product inputs, source checks, and code. It supports within-session method inspection. It does not test clinical benefit or performance in a new session.

SEEG · electrical stimulation · bipolar readout · response shape · signed gain · within-session reliability · human limbic system