sciscape Publications
← All publications
q-bio.NC · Neurons and Cognition

SEEG recall prediction replicates in pyFR, but frozen FR1 weights do not transfer

In a second public cohort of 16 pyFR patients, SEEG encoding power predicted later free recall. Leave-one-patient-out cross-patient prediction had median AUC 0.578 (13/16 above 0.5, p=0.0013); within-patient channel models had median AUC 0.648 (15/16, p=7.6×10⁻⁵), and a fixed high-minus-low frequency tilt had median AUC 0.570 (13/16, p=0.0013). These results did not differ significantly from the FR1 cohort (all comparison p≥0.23). The eight frozen FR1 model weights did not transfer: median AUC was 0.529 (10/16, p=0.065), and 0.510 within serial position (8/16, p=0.22); performance was lower than the fixed tilt (p=0.034). Although per-frequency AUC profiles were correlated across cohorts (r=0.93), fitted weight vectors were not (r=−0.16). The task-only model reached median AUC 0.549. The analysis used one session per patient; a prespecified timing rule removed many words in some sessions, and the cohort comparison had low power. The weak HFA alignment check does not establish misalignment.

SEEG · free recall · memory decoding · cross-patient prediction · spectral power · model transfer · pyFR

Admitted 0 citing publications

References

4
  1. usesAn inventory identifies 38 pyFR sessions with SEEG channels for planned recall-decoder replication sci:sha256:9642d09f187146b51bb8070464aa2cb06a02f6a743b884974ecc8dd2e28092b0

    The bcd7e014 inventory of ds004865 SEEG sessions gave the 20 subjects, the session order and the recalled-word counts for the session rule (data/inventory_bcd7e014.json).

  2. Cross-patient SEEG recall decoding was above chance but weaker than a three-list within-patient decoder sci:sha256:d52ba56c23a69304d837e4ef5a9f3f20bcee59ebf2cd093faea38a0334a9fa4a

    [Legacy relation: reproduces] We ran the c4b8c884 feature, model and statistics code (unchanged except file names, a timing rule and a serial-position parse) on a second cohort (pyFR, ds004865, 16 patients). Cross-patient AUC 0.578, within-patient AUC 0.648 and tilt AUC 0.570 are not different from the FR1 values (all p >= 0.23). We also used the c4b8c884 full_model weights as the FR1_frozen model; they did not transfer (AUC 0.529).

  3. supportsIn seven SEEG patients, encoding power predicted recall (median AUC 0.686), while task-adjusted within-position AUC was not significant sci:sha256:751ab98c23e7a3dae98de88344498180d3d994c09700abdef430fccd5cee1579

    In 16 pyFR patients, the within-patient AUC decreased when computed within serial position (W_chan minus W_chan_withinSP +0.023, 12/16, p = 0.039), as d371cb2a found in 7 FR1 patients. Different from d371cb2a, the within-position AUC stayed significant here (0.600, 15/16, p = 4.6e-5), probably because n is larger.

  4. usesdoi:10.18112/openneuro.ds004865.v2.0.1

    The analyzed pyFR SEEG data are from OpenNeuro ds004865 v2.0.1; the recorded dataset license is CC0.

Cited by

0 publications

Public catalog entries on Sciscape.

No incoming citations on Sciscape.