Calibration shape selection gave higher mean site accuracy than random four-site sets for 6 mA pulses in five SEEG sessions at a 15 mm distance limit

We tested whether calibration response shapes can select four stimulation sites that are easier to identify from later pulses at those sites. We used public SEEG data from five patients in OpenNeuro ds004457 version 1.0.2. Each patient had one fixed recording pair. The main test used 6 mA pulses and a minimum 15 mm distance rule. There were 146 candidate sites. The selected sets used 100 calibration pulses and 107 test pulses. Mean site accuracy was 67.7%, compared with 45.1% for 1,000 random four-site sets per patient. The mean gain was 22.6 percentage points. It was positive in all five patients. The descriptive patient bootstrap interval was 10.5 to 35.6 points. The exact two-sided patient sign-flip result was p = 0.0625. Shape selection did not clearly exceed selection by response energy. A 30 mm distance rule and removal of trials with specific source notes weakened the result. Pulses were split within each site. No selected test set occurred fully after all calibration pulses. This is an offline test of recorded response separation. It does not establish perception, neural origin of all signal information, safety, or online interface performance. We verified all raw file hashes, rebuilt the trial arrays, and independently checked all saved predictions. The package preserves code, data bins, source versions, results, figures, audit arrays, and raw-data retrieval instructions.

SEEG · intracranial EEG · electrical stimulation · site identification · waveform shape · calibration · site selection · response energy · ventral temporal cortex · brain-computer interface · epilepsy