The cohorts
Two independent resting-state EEG datasets
Both are public (CC0), eyes-open resting-state recordings, harmonized to 31 shared 10–20 electrodes. They differ in hardware, sampling rate, country, and class balance — exactly the differences a transferable biomarker must survive.
| dataset | role | n | PD / HC | age (range) | male | features |
|---|---|---|---|---|---|---|
| ds007526 | development | 133 | 106 / 27 | 65 (39–84) | 59% | 210 |
| ds002778 | external validation | 30 | 15 / 15 | 63 (47–82) | 47% | 210 |
Counts are after quality control. The development cohort is imbalanced (~4:1 PD:HC) and its patients are older and more male than controls — a confound quantified on the Results page. The external cohort is age/sex-balanced, which is why the demographic shortcut fails to transfer.