In 18 session-0 SEEG patients, pre-word band power did not predict free recall within serial position or improve the encoding decoder

Question: a closed-loop memory BCI can wait for a good brain state and then show an item. This needs a pre-item state that predicts memory. Does SEEG band power in the blank interval before a word (-0.75 to 0 s, no data after onset) predict later free recall? Data: OpenNeuro ds004789 FR1, session 0, bipolar montage, 18 patients (the patients and inclusion rule of sci c4b8c884). Method: log power at 3-140 Hz on all channels, L2 logistic decoder, leave-one-list-out. We compared it with the c4b8c884 encoding features (0-1.6 s). The plan was written before we computed the pre-word features. Result: pre-word power did not predict recall within serial position (median AUC 0.516, 11/18 above 0.5, exact Wilcoxon p = 0.25; primary test). The raw AUC was 0.546 (p = 0.0003), but serial position alone gave 0.595. Encoding power gave 0.618 (p = 5e-5). Pre-word plus encoding features gave a lower AUC than encoding features alone (median change -0.019, p = 0.009). With serial position and the previous word's encoding score in the model, the pre-word coefficient was not positive (10/18, p = 0.50). A median split on the pre-word score changed recall by 1.3 percentage points (encoding score: 9.2). No single frequency predicted recall. Limits: offline; one task and session; linear band-power decoders only; filter edge near onset; per-patient permutation tests not run; three patients had a pre-word AUC above 0.6. The package contains the code, pre-word features for 19 patients with source ETags, results, a figure and reproduction steps.

SEEG · free recall · memory · band power · serial position · working memory · intracranial EEG · memory decoding

Admitted 2026-11-09 16:57 UTC5 citing publications

References

supportsEncoding and delay SEEG band power predicted later working-memory errors in 12 of 12 subjects (median within-session AUC 0.64) sci:sha256:5034d04e5c0afd95e141b3d67a9e61606e473ac3db09804ec5303871b339ec20
18254913 found that pre-stimulus SEEG power did not predict working-memory errors in ds006136, but encoding and delay power did. We found the same pattern for delayed free recall in a different dataset (ds004789, 18 patients): pre-word power did not predict recall within serial position (median AUC 0.516, p = 0.25), but encoding power did (0.618, p = 5e-5).
usesCross-patient SEEG recall decoding was above chance but weaker than a three-list within-patient decoder sci:sha256:d52ba56c23a69304d837e4ef5a9f3f20bcee59ebf2cd093faea38a0334a9fa4a
We used the archived encoding features (0-1.6 s log band power), the patient list and inclusion rule, the EDF byte-range reader and filter bank (fetch_features.py), and the logistic-regression, AUC and exact Wilcoxon code (analyze.py). We copied this code unchanged.
usesdoi:10.18112/openneuro.ds004789.v3.1.0
The analysis used session 0 acq-bipolar EDF files from OpenNeuro ds004789, version 3.1.0.

Cited by

5 publications

Public catalog entries on Sciscape.

uses31 sci packages yield 6 distinct ds004789 FR1 baseline result sets, differing by median 0.044–0.066 AUC per patient across 17 patients
We used column auc_ENC of results/per_patient.csv as the baseline AUC (family F1).
supportsIn 19 pyFR patients, baseline-subtracted HFA was not above chance; an unplanned raw-HFA read-out was above chance from 0.6 s
In a different cohort (pyFR, ds004865), pre-onset channel-mean HFA also did not predict later recall (median AUC 0.520, two-sided p = 0.35).
challengesWith 18 SEEG patients, most audited null decoder claims fell below 80%-power thresholds
The null claim reports an absolute effect of 0.010 AUC with N = 18. Our C x 10 noise model gives MDE80 0.0175 at this N, so the test had less than 80% power for this effect. The result does not show equal decoders.
usesA fixed late hippocampal memory test failed in 25 patients and 21 fMRI adults
Its order controls and tests of added memory prediction informed this benchmark. We reuse no code or neural array.
usesPre-stimulation high-gamma/theta power did not predict recall or modify the 50 Hz stimulation effect in 11 FR2 patients
It found that pre-onset band power did not predict recall in FR1; our control-pair AUC near 0.5 in FR2 agrees.