Late SEEG shape gave 27.7 percent mean balanced shaft accuracy without shared stimulation contacts in five patients

We tested prediction of the stimulated shaft from late SEEG shape at separate stimulating contacts. We used OpenNeuro ds004457 version 1.0.2, five patients, 136 test pairs, and 16 receiving contacts. The model used means of five pulses at known receivers within each patient. Training and test pairs shared no stimulating contact or source pulse. Each contact waveform was centred and divided by its norm. The fixed late window used centres from 100 to 280 ms. Mean balanced accuracy was 27.7 percent after stimulation and 16.9 percent before stimulation. All five patient differences were positive. The two-sided patient sign-flip test gave p = 0.0625 under its symmetry and independence assumptions. This test cannot pass 0.05 with five patients. The amplitude model gave 31.4 percent after stimulation. The study did not establish a shape gain over amplitude. An independent calculation checked all 2,504 predictions. A raw reconstruction matched every input array, including 3,477,760 waveform values. The package preserves the local plan, data, source locks, code, folds, uncertainty draws, figures, checks, and reproduction commands. This is a secondary analysis of the same clinical cohort used in earlier work.

SEEG · electrical stimulation · waveform shape · shaft classification · contact-separated prediction · balanced accuracy