Visual broadband detectability was similar for mapped V1–V3 SEEG and ECoG contacts, but varied with contact location

We compared visually driven 70–200 Hz broadband power in 30 SEEG depth contacts and 128 ECoG surface contacts from nine patients using the Groen et al. temporal-pattern task derivative in OpenNeuro ds004194. Seven patients had both electrode types. Pooled median single-trial detection AUCs were similar in Benson-atlas V1–V3 contacts (0.82 for SEEG, n=12; 0.83 for ECoG, n=41; electrode-level Mann–Whitney p=0.53). Across all areas, the mean within-patient SEEG-minus-ECoG AUC difference was −0.127 (7 patients; permutation p=0.008); SEEG AUC was lower in six patients. In those six, none of 13 SEEG contacts was responsive (AUC≥0.65), and eight were outside the V1–V3 maps. The V1–V3 within-patient test included three patients and was not significant (difference −0.17, p=0.63). In one patient, all 17 SEEG contacts in V2, V3, V3a/b, hV4, LO1 and VO2 were responsive (AUC 0.68–0.95). Both modalities showed compressive temporal summation (k<1 in 11/12 SEEG and 63/65 ECoG fits). All fitted SEEG contacts came from one patient, so the larger SEEG exponent (median 0.46 vs 0.26; electrode-level p=0.001) cannot be interpreted as a modality effect. The data do not separate electrode type from contact location; atlas labels were not patient-specific functional maps. These results suggest contact location is the more probable predictor of visual read-out quality in this dataset.

visual cortex · SEEG · ECoG · broadband power · single-trial detection · temporal summation · brain-computer interface · electrocorticography