One repeated SEEG site retained identification accuracy across 50 minutes, while cross-site timing effects remained anatomically confounded

We tested whether block timing can affect single-pulse stimulation-site identification from SEEG responses. In sub-1, a nearest-centroid decoder identified pulses among 39 sites. For the clean repeated site LB1-LB2, with 9 and 14 pulses in blocks 50 minutes apart and 42 intervening blocks, other-block templates performed as well as same-block templates (accuracy 0.42–0.52 versus 0.40–0.50). Across the 50 pulses with templates of at least 3 pulses, accuracy was 0.28 for other-block and 0.24 for same-block templates (paired sign-flip p = 0.30). The source summary also reports all 61 tested pulses, including the one-pulse template: same-block accuracy 0.300 and other-block accuracy 0.235 (two-sided sign-flip p = 0.138). A template based on one pulse failed. Across site pairs in two patients, response distance increased with time between blocks, including after controlling for Euclidean midpoint distance and among different-shaft pairs. However, stimulation order followed shafts and anatomy, so this analysis cannot separate drift from anatomical effects. Within-block chronological splits showed no consistent effect. The direct repeated-site test supports template stability for one site in one patient; it does not establish general stability. Pulses are not independent, p values treat pulses as units, features include baseline removal, and Euclidean midpoint distance was the only spatial control. The results support interleaved or randomized stimulation order for BCI studies.

SEEG · electrical stimulation · single-pulse stimulation · stimulation-site identification · evoked responses · block drift · template stability · brain-computer interface