Selection from three windows and one early window gave the same mean lowest site accuracy in five offline SEEG sessions

We tested site selection with public SEEG data from five patient sessions in OpenNeuro ds004457 version 1.0.2. Each set had eight stimulation sites. We fitted readers with the first four eligible pulses per candidate site and tested the last four. Every response window lasted 80 ms. A greedy rule selected sites with large distances across starts of 20, 60, and 100 ms. The comparison rule used only the 20 ms start. Both rules gave 58.125% mean lowest test accuracy across the three starts. The mean gain was zero points. The descriptive 95% patient bootstrap interval was -10 to 8.75 points. Three patients improved and two worsened. The rule for three windows increased the minimum calibration distance in all five sessions. This did not give a mean test gain. The result does not prove equivalence. This offline test used known pulse times and separate readers for each start. It did not test amplifier recovery noise or prospective use. The package preserves the local plan, compact data, raw file hashes, readers, predictions, figures, code, and successful raw rebuild and replay checks.

stereoelectroencephalography · site selection · electrical stimulation · brain-computer interface · response decoding · window selection · offline evaluation