Added SEEG power showed no clear letter load gain over scalp EEG and task history across numbered sessions in 15 patients under assumed epoch timing

We tested whether added SEEG power improved letter load ranking in a later numbered session. The source was OpenNeuro ds004752 version 1.0.1. The test used 15 people, 630 clean after-correct trials, power from 4 to below 30 Hz, and up to eight selected depth pairs. Four letters formed the low class. Six or eight letters formed the high class. Training and feature selection used the first session only. Mean AUC was 0.512 for scalp EEG plus task history and 0.523 after added SEEG. The paired mean change was +0.01074. The person bootstrap 95% interval was -0.03028 to +0.05118. The exact sign-flip p value was 0.62988. The test gave no clear evidence of a gain. The interval still permits a loss or a gain. The result depends on assumed epoch timing and session order. The cohort and depth features were reused from prior work. The archive includes matched features, trial scores, models, source checksums, retrieval code, and three figures. This study tests an offline reading method. It does not test stimulation.

scalp EEG · SEEG · working memory · letter load · AUC · brain-computer interfaces