Two calibration pulses gave a mean score gain of 0.062 bits over eight sites for 20 to 100 ms SEEG responses in five sessions

We tested whether two separate calibration pulses could choose a site count that improves SEEG response reading. We used five public sessions from OpenNeuro ds004457 version 1.0.2. There were 129 eligible sites with at least ten valid pulses each. Three pulses fitted site means and a fixed covariance metric. Pulse 4 fitted a shared probability scale. Pulses 5 and 6 chose among 2, 4, 8, 16, and all sites. The last four pulses per site supplied the test. The main window was 20 to 100 ms after the pulse. Calibration chose 8, 8, 16, 16, and 16 sites. Mean test scores were 2.703 bits per pulse for this choice and 2.641 for fixed eight-site sets. The mean gain was 0.062 bits per pulse. The descriptive 95% patient bootstrap interval was -0.258 to 0.389 bits. Two sessions improved, one declined, and two tied. These data do not show a clear gain from the choice rule. The score includes label entropy and probability error. It is not brain-writing capacity. The split was forward within sites, but not across the whole session. The patients and test pulses were reused from prior work. The package preserves figures, derived voltage data, code, pulse roles, source hashes, and a raw retrieval recipe. All 110 source files passed size and hash checks. A raw rebuild matched the inputs exactly. An independent probability check and an exact repeat analysis passed.

stereoelectroencephalography · SEEG · electrical stimulation · response decoding · probability log score · site selection · calibration · human neuroscience