A six-patient pilot found no corrected hippocampal power effect of working-memory load in SEEG

This six-patient pilot examined whether bipolar hippocampal SEEG band power during verbal working-memory maintenance changed with set size (4, 6, or 8 consonants), and whether power predicted response time (RT) within set size. We used ses-01 from six of 15 patients in OpenNeuro ds004752 and analyzed correct, artifact-free trials in the 4.0–6.0 s window, the last 2 s of maintenance. Welch log power was measured in delta, theta, alpha, beta, low-gamma, and high-gamma bands. RT increased with load in all six patients (exact sign-flip p=0.031). Hippocampal power showed no significant load effect in any band after Holm correction (all adjusted p=1.0); theta and alpha were nominally higher at set size 8 than 4 by 0.53 and 0.33 dB, respectively. Higher within-set-size hippocampal alpha power was associated with slower RT in all six patients (uncorrected p=0.031; adjusted p=0.19). The amygdala comparison showed no corrected load or RT effects. This pilot is underpowered: with six subjects, the minimum exact two-sided sign-flip p is 0.031, so it cannot survive correction across six bands. A null result is not evidence of no effect. Three additional subjects failed an EDF alignment assertion, and six others were not processed in the reported run. Atlas labels were used without gray/white-matter or seizure-onset-side checks; several patients have hippocampal sclerosis. Only one session per patient and the final 2 s of maintenance were analyzed, with no interictal epileptiform discharge detection or baseline normalization. The package preserves the numpy-only range-request code, feature files, logs, and result tables. The source used a live OpenNeuro S3 URL; the exact dataset release is uncertain.

intracranial EEG · working memory · hippocampus · response time · SEEG · spectral power · pilot study