ECoG had higher best-contact visual-onset AUC than SEEG in 8 patients, but not after contact-count matching

In 8 patients with both SEEG depth and ECoG surface contacts, we compared single-trial visual-onset detection using 70–170 Hz power. The best ECoG contact had higher AUC than the best SEEG contact in 7 of 8 patients (median 0.853 vs 0.633; exact sign-flip p = 0.031). The mean AUC of the three best contacts was also lower for SEEG in 7 of 8 patients (median 0.615 vs 0.782; p = 0.047). ECoG had more contacts. After matching the ECoG contact count to each SEEG set by random subsets, the expected best ECoG AUC remained higher in 7 of 8 patients, but the mean difference was smaller (-0.096) and not significant (p = 0.19). The fraction of responsive contacts did not differ significantly (p = 1.0). In one patient, 20 of 25 SEEG contacts were responsive and the best SEEG AUC was 0.952, showing that SEEG can detect visual onset well when contacts are in responsive visual cortex. Contact locations were selected for clinical needs and were not controlled; the findings describe these implants and do not establish the intrinsic quality of depth recording. The analysis included 8 patients, one task, 108 stimuli per contact, one power feature, and offline onset-versus-baseline detection; it did not decode stimuli. There was no correction across the four group tests.

electrocorticography · stereo-EEG · visual cortex · single-trial detection · high-frequency power · AUC