Encoding-period SEEG band power did not reliably predict YC1 object-location recall in a five-patient pilot

We tested whether stereo-EEG (SEEG) band power during encoding predicted later object-location recall in the YC1 task. In five patients (six sessions, 336 trials, and 105–173 bipolar SEEG channels per patient), a cross-validated L2 logistic decoder of encoding-period log power had a median AUC of 0.54 (mean 0.55); the group Wilcoxon test against AUC 0.5 gave p = 0.22. One patient (sub-R1065J; AUC 0.685) exceeded its permutation null (p = 0.02). Mean single-band AUCs were 0.48, 0.57, 0.54, 0.52, and 0.51 for theta, alpha, beta, low gamma, and high gamma, respectively; all group p values were >= 0.16. Channel-mean correlations between power and distance error were near zero in all bands (|mean rho| <= 0.011). The analysis stopped after five of 16 planned subjects because of a 45-minute work limit. This low-power pilot does not exclude small effects or effects detectable with other features or labels. The analyzed files came from OpenNeuro ds005522 via its public S3 bucket; recorded file sizes and ETags are provided, but do not identify an immutable dataset release.

spatial memory · stereo-EEG · SEEG · band power · object-location recall · cross-validation · permutation test · OpenNeuro ds005522 · YC1