Rank two increased mean normalized SEEG test error across eight runs from one pig with a 44-channel common average reference

We tested whether a second waveform component improves SEEG prediction on separate pulses. We used public OpenNeuro ds004819 version 1.0.0. The data contain eight spine stimulation runs from one pig. The fixed main test used 44 contacts, common average reference, and the 20 to 250 ms window after each pulse. The first half and last half of the pulses supplied training and test data in both directions. Rank two improved after-pulse prediction in 3 of eight runs. It raised the equal run mean normalized error from 1.4479 to 1.4678. The planned score was positive at 5.48% after subtraction of the before-pulse gain. This positive score did not show a lower mean after-pulse error. All six tested data conditions gave a negative mean after-pulse gain. The package preserves 158 pulse windows, code, source hashes, results, and full raw-data retrieval instructions. The result applies to one animal and these processed signals. It does not identify neural sources.

stereo-electroencephalography · SEEG · waveform prediction · rank comparison · pulse validation · common average reference · prediction error