A second SEEG shaft was recruited a median 3 s after the first pair in 76 seizures from 22 patients (ds004100; 1-s HF power, z >= 3 for 3 s)

We measured how fast focal seizures spread in SEEG. This time limits a responsive-stimulation BCI. We used archived 1-s features of 79 seizures from 22 patients (OpenNeuro ds004100, HUP). A bipolar pair was recruited when its 20-100 Hz power was 3 SD above its pre-ictal baseline for 3 consecutive 1-s windows. A sham window before the onset mark gave the noise rate. Recruitment occurred in 76 seizures. The patient-median fraction of recruited pairs was 29.9%; the sham fraction was 0.27% (p < 1e-4). From the first recruited pair, activity reached a pair on a different shaft after a patient median of 3.0 s (IQR 2.0-5.4 s). In 41% of seizures this occurred within 2 s. 10% of pairs were recruited after 6.5 s. 25% of pairs needed 13 s or more (41% censored at the end of the data). Line length gave the same result (3.5 s to a second shaft). Thresholds z >= 2 and z >= 5 gave 3.0 s and 3.75 s. Removal of seizures with recruitment at the search start did not change the result. Clinical SOZ pairs were recruited earlier than other pairs (AUC 0.69, 17 of 21 patients above 0.5; sham AUC 0.49, p = 0.003). But the first recruited pairs included a SOZ pair in only 51% of seizures. A detector with 3 s of confirmation thus confirms a seizure at about the time it reaches a second shaft. To act before this spread, a BCI must detect faster or stimulate more than the first shaft. Limits: one center, clinical labels, threshold rule, 1-s resolution, data end 19 s after the onset mark, no anatomical distances.

seizure spread · stereo-EEG · SEEG · responsive neurostimulation · brain-computer interface · high-frequency power · seizure onset zone · OpenNeuro ds004100