Full baseline correction increased mean error for centered early gamma maps in 32 SEEG patients with two seizures each and common average reference

We tested whether baseline power changes improve the transfer of early seizure maps between two recordings from one patient. We used public OpenNeuro ds004100 version 1.1.3 and a fixed group of 32 SEEG patients. The test had 64 seizures and 3,212 matched contact anchors. Gamma power used 30 to 57 Hz and 63 to 80 Hz. Baseline power used -30 to -10 seconds relative to the supplied onset. Early power used 0 to 10 seconds. We used common average reference and removed each log map's contact mean. Full correction added the complete baseline map change to the first early map. Equal-patient mean squared error increased from 8.20 to 10.90 dB squared. The increase was 2.71 dB squared, with a 95 percent patient bootstrap interval of 0.92 to 4.61. This interval assumes independent patients. Error increased in 26 of 32 patients. All eleven planned method checks had higher mean error. We give an exact condition for benefit and a descriptive test of a smaller correction coefficient. Clinical time order was not established. The result concerns map transfer, not clinical localization. A fresh raw audit passed all 64 recordings and 757,770 stored power values. The package preserves input features, pinned retrieval records, code, predictions, results, proof, and reproduction instructions.

stereoelectroencephalography · seizure · gamma power · baseline correction · map transfer · intracranial EEG