Software Credentialed Access
Optimal spindle detection parameters for predicting cognitive performance
Noor Adra , Haoqi Sun , Wolfgang Ganglberger , Elissa Ye , Lena L. Dümmer , Ryan A. Tesh , Mike Westmeijer , Madalena D. S. Cardoso , Erin Kitchener , An Ouyang , Joel Salinas , Jonathan Rosand , Sydney Cash , Robert Thomas , M. Brandon Westover
Published: July 10, 2026. Version: 1.0.0
When using this resource, please cite:
(show more options)
Adra, N., Sun, H., Ganglberger, W., Ye, E., Dümmer, L. L., Tesh, R. A., Westmeijer, M., Cardoso, M. D. S., Kitchener, E., Ouyang, A., Salinas, J., Rosand, J., Cash, S., Thomas, R., & Westover, M. B. (2026). Optimal spindle detection parameters for predicting cognitive performance (version 1.0.0). Brain Data Science Platform. https://doi.org/10.60508/f4pp-1g21.
Abstract
Study objectives. Sleep spindles are candidate biomarkers of cognition, but detected spindle characteristics depend strongly on detector settings. We asked which spindle-detection parameters yield spindle features that best predict cognitive performance.
Methods. Across a large grid of spindle-detector parameter settings, spindle features were extracted from overnight EEG and related to NIH Toolbox cognition scores using cross-validated regression, identifying the parameter settings that maximize prediction of cognitive performance.
Results. The optimal detection parameters and their predictive performance are reproduced from the committed data; fluid-composite cognition was predicted with correlation on the order of r ~ 0.5.
This project provides the de-identified spindle-feature data (BDSP-linked), the cognition data, and the code to reproduce the parameter optimization.
Background
Sleep spindles are proposed biomarkers of cognition, but which spindle-detector settings produce the most cognition-predictive features is unclear. This project releases the data and code from a systematic optimization over spindle-detection parameters.Software Description
De-identified spindle-feature files per subject and parameter set (data/bridge_spindles/, keyed by BDSP/SPNDL IDs), a de-identified sibling cohort (data/admci/), the cognition table (age_cog_data.csv, A### IDs), a de-identified join key (spindles_id_pairs_deidentified.csv linking A### to spindle IDs, no names), and the Python/LUNA analysis code. No names, MRNs, or absolute dates.Technical Implementation
Spindle features were extracted from overnight EEG across a large grid of detector parameter settings and related to NIH Toolbox cognition via cross-validated regression to identify the parameters that maximize prediction of cognitive performance.Installation and Requirements
Python 3.9+ with pandas, numpy, scipy, scikit-learn, matplotlib. See REPRODUCE.md; join cognition to spindle features via spindles_id_pairs_deidentified.csv.Usage Notes
Reproduce the parameter optimization with the committed de-identified spindle features + cognition, joined by spindles_id_pairs_deidentified.csv (A### <-> BDSPID/SPNDL). The committed result tables reproduce the paper's headline numbers.Release Notes
First public release: de-identified spindle-feature + cognition data + parameter-optimization code.Ethics
De-identified data under IRB approval; identifiers removed, dates shifted, BDSP IDs retained; SHHS/NSRR cohort excluded.Acknowledgements
We thank the participants and the BDSP.Conflicts of Interest
See the associated publication (Sleep 2022;45(4):zsac001).References
- Adra N, Sun H, Ganglberger W, Ye EM, Dummer LW, Tesh RA, Westmeijer M, Cardoso MDS, Kitchener E, Ouyang A, Salinas J, Rosand J, Cash SS, Thomas RJ, Westover MB. Optimal spindle detection parameters for predicting cognitive performance. Sleep. 2022;45(4):zsac001. PMID: 34984446.
Access
Access Policy:
Only credentialed users who sign the DUA can access the files.
License (for files):
BDSP Credentialed Health Data License 1.5.0
Data Use Agreement:
BDSP Credentialed Health Data Use Agreement
Required training:
Discovery
Corresponding Author
Files
- be a credentialed user
- sign the data use agreement for the project