Four sites selected from SEEG shapes at 30–100 ms gave higher mean pulse identity accuracy than random sets within five patient sessions

We tested whether calibration SEEG shapes could select four stimulation sites that were easier to identify in later pulses. We used public OpenNeuro ds004457, version 1.0.2, with 1,923 pulses and 185 sites in five patients. Each patient had one fixed recording contact pair. The first half of pulses at each site supplied calibration. A greedy rule used waveform differences at 30–100 ms. The later half supplied the test. Mean site recall was 86.2% for selected sets and 51.1% for random four-site sets. Each patient showed a gain. The mean gain was 35.1 percentage points. The exact two-sided patient sign-flip test gave p = 0.0625. The same selected sites gave 34.3% recall before the pulse, above the 25% chance value. The test holds out pulses within stimulation blocks, not new sessions. It does not establish a percept or useful brain input. The package contains four figures, signal bins, code, predictions, and source hashes.

stereoelectroencephalography · electrical stimulation · brain-computer interface · site identification · signal classification · calibration