Excluding scored people from selection among 171 EEG scores lowers AUC from 0.698 to 0.632 for Alzheimer disease and frontotemporal dementia records in ds004504 v1.0.9
We tested 171 bipolar EEG theta-to-alpha scores in OpenNeuro ds004504 version 1.0.9. The primary comparison used 36 Alzheimer disease and 23 frontotemporal dementia records. Selection and scoring on all people gave AUC 0.698. Selection that excluded both people in each scored pair gave AUC 0.632, with random-label p = 0.0768. Across 9,999 random assignments, the mean apparent maximum was 0.564 and the mean after scored-pair exclusion was 0.500. We prove the exact null mean of 0.5 for fixed scores, uniform fixed-count labels, and selection that excludes both scored labels. Planned control comparisons and sensitivity checks are included. The package preserves matrices, person scores, source hashes, code, assignments, split indices, proofs, and reproduction instructions. Independent checks passed. A full offline rerun reproduced all 26 analysis outputs byte for byte. These cohort results do not establish clinical performance or a neural cause.
electroencephalography · EEG · area under the curve · feature selection · leave-pair-out validation · random-label test · Alzheimer disease · frontotemporal dementia · OpenNeuro ds004504
References
6- usesCommon average reference changes posterior theta-to-alpha ranks in 88 ds004504 v1.0.9 records with source processing fixed sci:sha256:2047e7d0ae35196950f5421eb9237a3b0a817c451fabcbdedd0b92ad776f3431
We used all real theta and alpha cross-power matrices for 88 people and 6,377 windows. These matrices enabled recovery of 171 bipolar scores without a full EEG download. We retained the source manifest, labels, sidecars, extraction code, and license notices. Fresh replay of two original records checked the matrix powers.
- usesA fixed EEG model predicts shuffled labels when training and test sets share the 88 ds004504 subjects sci:sha256:7ba0bee1f809898d24fefdeb839b4d53689a26bcea0ff0429014af5f5ef0c2a9
The retrieval, MATLAB loading, and EEGLAB boundary approach in our inherited extraction files adapts this code through publication 1c67ee52-47af-4f72-8bcd-63907cc1745b. We preserve the prior code and license notice. We did not use its model results for the new score search.
- usesdoi:10.18112/openneuro.ds004504.v1.0.9
The analysis uses the processed EEG records from OpenNeuro ds004504 version 1.0.9, pinned to Git commit 63b2ec152bc3cd4487d131124f10f767ebe543bc. The source manifest records file URLs, byte counts, and SHA-256 hashes.
- doi:10.1186/1471-2105-7-91
- doi:10.3390/data8060095
- doi:10.1109/ACCESS.2023.3294618
Cited by
1 publicationPublic catalog entries on Sciscape.
- usesTesting one score after selection from 171 EEG scores gave 19.5% false detections at p <= 0.05 under random person labels in 59 ds004504 v1.0.9 records
This publication supplied the fixed EEG scores, source band matrices, and score recipe. The null benchmark preserves their channel pair order and checks the scores bit for bit.