In two analyzed patients, visual broadband responses varied by contact and patient; electrode-type effects remain unresolved

This incomplete analysis compared visual broadband responses at clinical SEEG depth and ECoG surface contacts during a 12-class spatial-pattern task in OpenNeuro ds004194. Eight patients were planned; features were extracted for four and the partial log reports results for two. In p01 (8 SEEG, 52 ECoG contacts), almost no contacts were responsive (0% vs 2%); matched detection AUC was 0.519 vs 0.516 and identity accuracy was 0.049 vs 0.083 (chance 0.083). In p05 (25 SEEG, 146 ECoG), 96% of SEEG and 11% of ECoG contacts were responsive (BH q < 0.05); at 16 matched contacts, detection AUC was 0.936 vs 0.700 and identity accuracy was 0.179 vs 0.116. All-contact detection AUC was 0.951 vs 0.882; all-contact identity accuracy was 0.167 for both types (binomial p = 0.0009 for each). No group test was done. Contact locations were not anatomically matched, and the common average reference was computed across all channels by the dataset authors. The p05 difference therefore cannot be attributed to electrode type. The archived consolidated results file contains p01 only, whereas the partial log also contains p05; this inconsistency is documented in the full text.

intracranial EEG · SEEG · ECoG · visual cortex · broadband power · single-trial decoding · incomplete analysis