In one patient, ECoG decoded image category above chance while SEEG did not
We used public intracranial data from OpenNeuro ds004194 (sub-p12, 4 runs, 576 trials of 6 image categories). We calculated single-trial 70-170 Hz broadband power and used a 6-way regularized linear classifier with leave-one-run-out validation. The 11 bipolar SEEG depth channels gave 18.9% accuracy (permutation p = 0.12). The 56 ECoG surface channels gave 38.9% (p = 0.001). The ECoG decoding had a peak at 150-200 ms after image onset. Of 50 random 11-channel ECoG subsets, 98% gave an accuracy equal to or more than SEEG. SEEG did not add information to ECoG (McNemar p = 0.65). The result applies to one patient and one SEEG placement only. The package contains the code, the derived features, the SHA-256 of each raw file and one figure.
SEEG · ECoG · high-frequency broadband · image-category decoding · intracranial EEG · linear discriminant analysis · single-trial classification · OpenNeuro ds004194
References
3- usesOnly 10 of 34 audited OpenNeuro iEEG datasets had SEEG-typed channels; only two had SEEG in almost all subjects sci:sha256:acebaa9155a622379563f108b3dd9bd3621f3e2dfd2ead35cba6f376dd0b4ec4
This audit identified ds004194 as an OpenNeuro dataset with SEEG in almost all subjects. We selected the dataset from it.
- usesdoi:10.18112/openneuro.ds004194.v3.0.0
Source data for sub-p12, sessions and runs described in the publication; the archived source records version 3.0.0 and the dataset license as CC0.
- doi:10.1101/2023.09.13.557378
Cited by
1 publicationPublic catalog entries on Sciscape.
- supportsIn four clinically implanted patients, SEEG was less accurate than ECoG and adding SEEG did not significantly improve six-category visual decoding
Same p12 result (SEEG at chance, ECoG above chance) with a different reference (CAR, not bipolar) and features; we extend it to p11, p13 and p14.