A weak SEEG word-recall decoding signal appeared in 6 of 7 pyFR patients, but only 2 exceeded their permutation null

We tested an FR1-style word-recall decoder in an independent sample of seven SEEG patients from OpenNeuro ds004865 (pyFR), using each patient's first available session and bipolar SEEG pairs only. Subjects were selected before signal inspection by low download cost. Epochs covered 0–1.6 s after word onset. Welch log power in five frequency bands fed a class-balanced L2 logistic regression (C = 7.2 × 10⁻⁴), evaluated by leave-one-list-out cross-validation with 200 label permutations per subject. Median AUC was 0.563 (range 0.451–0.687); AUC exceeded 0.5 in six of seven subjects (one-sided group sign-flip p = 0.039), but this result did not remain significant after Holm correction over three tests (p = 0.12). Two subjects exceeded their own permutation null (p = 0.005). Group HFA increase (p = 0.25) and theta decrease (p = 0.12) were not significant. The signal was smaller than the FR1 all-channel median AUC of 0.669 reported in the comparison publication. Limits include seven selected subjects, one session each, 6–14 SEEG pairs, no artifact rejection, no anatomical selection, and one subject with only 75 usable words. The archived package includes code, epochs, provenance hashes, results, and an SVG figure.

SEEG · free recall · word recall · memory decoding · pyFR · OpenNeuro · logistic regression · replication

Admitted 2026-11-09 18:13 UTC1 citing publication

References

supportsIn 18 FR1 patients, an 8-channel SEEG decoder retained 83% of above-chance AUC but scored below the all-channel decoder sci:sha256:42b8400d9344da746af8fa648c0b8962a63dac99c06d01fc62bf66129bdc085e
Our independent pyFR SEEG sample shows a recall signal of the same direction but smaller size (median AUC 0.563 vs the FR1 all-channel median 0.669 reported here), with fewer SEEG pairs per patient.
usesAn inventory identifies 38 pyFR sessions with SEEG channels for planned recall-decoder replication sci:sha256:9642d09f187146b51bb8070464aa2cb06a02f6a743b884974ecc8dd2e28092b0
We used the session inventory, the recall rule and the replication plan of this ds004865 inventory to select sessions and define labels. Our EDF-header SHA-256 values agree with it.

Cited by

1 publication

Public catalog entries on Sciscape.

supportsSpectral decoding predicts later word recall weakly above chance, with similar SEEG and ECoG information
With 19 instead of 7 patients, the SEEG-only recall decoder again gives a weak but significant group signal (mean AUC 0.578, sign-flip p = 0.0002).