An FR1-trained SEEG model transferred to catFR1 recall (AUC 0.572), but serial position alone reached 0.574
A cross-patient model trained on uncategorized free recall (FR1) predicted later word recall in 31 patients who performed categorized free recall (catFR1): mean AUC 0.572 (95% CI 0.550 to 0.593; 27/31 above 0.5; Holm p = 0.00015). Its performance matched a leave-one-patient-out catFR1 model (mean difference -0.0004, 95% CI -0.015 to 0.015, p = 0.95), including a training-size-matched comparison. However, serial position alone gave mean AUC 0.574. After removing the serial-position mean from each neural feature, transfer AUC was 0.556 (95% CI 0.537 to 0.575; 26/31 above 0.5; Holm p = 0.00015). The 14 catFR1 patients not in the FR1 pool showed AUC 0.566 (13/14 above 0.5). The effect was small. Only session 0 was used, source features had no artifact rejection, and both tasks came from the same RAM project, hospitals, and pipeline; transfer to other sites or memory tasks was not tested. The full report preserves methods, all result tables, per-patient results, and limitations.
SEEG · free recall · memory · cross-task transfer · patient-independent model · band power · serial position · AUC · logistic regression · OpenNeuro
References
9- usesSample-timed tilt memory effect replicated, but HFA effect did not, in 15 new catFR1 SEEG subjects sci:sha256:0fdfc04b1336db0d3914dd8dfbfbda24b5ebd74512f05539fc3d2443e7ca44ab
Its catFR1 per-word band-power features are part of the catFR1 test pool.
- usesAcross 48 SEEG patients, a model trained on other patients predicted word recall above chance (mean AUC 0.549) sci:sha256:35ed98b4d206e63ff447d14f0eff28e0e24c0df8f2a7e87b805f25d49d653925
We reused its cross-subject decoder code unchanged (features, L2 logistic model, AUC, sign-flip and bootstrap) and its FR1 features as the training pool.
- usesIn 21 catFR1 SEEG subjects, EDF sample timing reveals a subsequent memory effect missed by onset timing sci:sha256:6054f551843194253a41e2d1f83a85704cd28044f964cf6bd91789bdd758fe07
Its catFR1 per-word band-power features are part of the catFR1 test pool.
- usesHFA, theta and tilt encoding effects replicated in 15 new catFR1 subjects, but the HFA effect weakened after serial-position controls sci:sha256:df479ea1af02a7c155bc6192f88da1b957ca0daac6ba4f0f0a384d7ec81f719e
Its catFR1 per-word band-power features are part of the catFR1 test pool.
- usesIn 51 pure-SEEG FR1 patients, sample-timed HFA decreased and theta increased with serial position sci:sha256:1da9d2dc0707f4d6cf26a2edf8d7e8f74cdb417581b20b70aafdf258db4aedc8
Its sample/fs FR1 per-word band-power features are part of the FR1 training pool (through 2f34e0d4).
- usesAssuming sample/fs marks EDF time, 21 pure-SEEG FR1 patients showed a subsequent memory effect hidden by onset timing sci:sha256:799055dd0e8b95498bb3ea5d9564670cf779d6970c9993bda2f0c41150b419e7
Its sample/fs FR1 per-word band-power features are part of the FR1 training pool (through 2f34e0d4).
- usesWith sample/fs epoch timing, the catFR1 null of 63055d12 changes to a subsequent memory effect (tilt +0.56, Holm p 0.0006) in 23 SEEG patients of ds004809, session 0 sci:sha256:f76bfd58d757a861b47c8d6b5ccc0494949caa735473fdc78db58946eb3a7cd5
Its catFR1 per-word band-power features are part of the catFR1 test pool.
- usesdoi:10.18112/openneuro.ds004789.v3.1.0
The FR1 training feature pool is identified as derived from OpenNeuro ds004789; saved version-pinned metadata records version 3.1.0 and CC0.
- usesdoi:10.18112/openneuro.ds004809.v2.2.0
The catFR1 test feature pool is identified as derived from OpenNeuro ds004809; saved version-pinned metadata records version 2.2.0 and CC0.
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.