Five early waveforms retained 61.2 percent of later SEEG energy with measured later coefficients in eight patient sessions

We tested a five-waveform representation bound for public SEEG in eight epilepsy patient sessions. The data were OpenNeuro ds004696 version 1.0.1. The test used 86 sites and 886 stimulation-to-readout pairs. Five early pulses defined each space. The final five pulses tested it. The 20 to 500 ms space retained 61.2 percent of later energy. A separate before-pulse space retained 48.9 percent. Both windows had 983 samples. Energies used early after-pulse normalization, with equal site and patient weights. The difference was 12.33 percentage points and was positive in all eight patients. The descriptive patient bootstrap interval was 9.65 to 15.34 points. However, the space retained only 47.4 percent of later after-pulse variation. Extra directions also retained background energy. Each projection coefficient used measured later voltage. Thus the result gives an optimistic representation bound, not a forecast available before stimulation. Independent SVD and event audits passed. Source-file hash records came from the prior extraction. The package preserves exact trial products, code, three figures, a reusable scoring tool, provenance, and reproduction instructions.

SEEG · stereoelectroencephalography · waveform representation · brain-computer interface · electrical stimulation · representation bound · principal components · OpenNeuro