Early energy reduced mean late energy prediction error at excluded stimulation pairs in seven SEEG patients after fixed source correction

We tested whether early response energy adds information about late response energy at an excluded stimulation pair. The data came from OpenNeuro ds004696 version 1.0.1. Seven patients supplied 80 groups and 1,511 contact-group rows. Predictors used five pulses. The target used five separate pulses. The base model included contact, pre-pulse energy, distance, and current. Adding early energy reduced mean patient squared prediction error by 18.01%. Six of seven patients improved. The approximate 95% patient interval was 0.20% to 35.83%. The exact two-sided sign flip probability was 0.03125. The gain was 17.21% when pulse roles were reversed. Direct late energy from the first pulse set gave a larger gain. The source reference used all exported trials. Thus the result is conditional on that fixed correction and recurring contacts within these patients. It does not establish neural causation or prediction in new patients. The package preserves reduced data, pinned source retrieval, all predictions, checks, and reproduction records. GPL source scripts are excluded, so exact public computational reproduction is not provided. A full source rebuild matched all sub-01 arrays. A second process reproduced all seven numerical output files exactly.

stereoelectroencephalography · SEEG · electrical stimulation · response energy · prediction · cross-validation · limbic system · epilepsy · OpenNeuro · late response