Added frontal and temporal power envelope correlations raised mean recall forecast error in fixed logistic readers for 11 SEEG patients under assumed source trial order

We tested whether early frontal and temporal power envelope correlations improve a later recall forecast. We used public bipolar SEEG from OpenNeuro ds006136 version 1.0.1. Eleven patients met a rule based on training coverage. There were 1,107 fit trials and 601 later test trials. Source trial IDs were assumed to give task order. A fixed logistic reader used context and regional power. A second reader added four beta and gamma envelope inputs. Mean patient Brier error rose from 0.175782 to 0.186691. Mean gain was -0.010909. Its descriptive 95% patient bootstrap interval was [-0.020696, -0.002778]. The exact two-sided patient sign-flip value was p = 0.054688. Four patients improved. Fixed penalty and shifted-envelope checks also had negative mean gains. The package adds 44,467 matched pair trials, all features, fitted readers, test forecasts, a digital signal example, three figures, source hashes, code, and replay instructions. Full source extraction and analysis repeated exactly. The cohort and later trials were used in prior studies. This is an exploratory extension, not independent validation. The result applies to the fixed inputs and readers. It does not test live control or brain stimulation.

stereoelectroencephalography · working memory · recall forecasting · power envelope correlation · Brier score · brain-computer interface