Why
Alzheimer's disease and related dementias (AD/ADRD) affect 7.2 million U.S. older adults, causing progressive, irreversible cognitive decline. There is hope that new treatment methods could slow or even halt progression, but current clinical assessments often can't detect the disease until significant damage has occurred.
The Challenge
The National Institute on Aging (NIA), part of the National Institutes of Health (NIH), ran the PREPARE Challenge (Pioneering Research for Early Prediction of Alzheimer's and Related Dementias EUREKA Challenge) to advance solutions for accurate, innovative, and representative early prediction of AD/ADRD.
This multi-year competition progressed through three phases, culminating in proof-of-concept approaches for real-world early AD/ADRD prediction and improving access to datasets and methods for researchers along the way:
Phase 1 addressed the foundational need of any model development effort: relevant, appropriately representative data. Solvers from across academia and industry were invited to find, curate, or contribute representative and open datasets relevant to early AD/ADRD prediction.
Phase 2 was a machine learning competition that used the data discovered in Phase 1 to advance algorithms and analytic approaches for early prediction of AD/ADRD, with an emphasis on explainability of predictions.
Phase 3 was a months-long refinement period for the top teams from Phase 2, who sourced new data and re-engineered their solutions to be more rigorous and generalizable in a real-world context, and who shared their methodological insights in non-technical presentations and reports to benefit the broader research community.
Prize overview
| Phase | Prizes Awarded |
|---|---|
| Phase 1 [Find IT!]: Data for Early Prediction | $190,000 |
| Phase 2 [Build IT!]: Algorithms and Approaches | $237,500 |
| Phases 3 [Put IT All Together!]: Proof of Principle Demonstration | $222,500 |
| Total | $650,000 |
The Results
Over 900 participants submitted solutions for open data and innovative methods, across all challenge phases. Ultimately, 10 teams from academia and industry developed proof-of-concept approaches across two complementary modeling tracks. Each track focused on a different source of early signals of cognitive decline. In the Social Determinants Track, solvers longitudinally predicted cognitive decline from survey data - capturing education, demographics, social experiences, employment, family history, and location - to predict cognitive decline over time. In the Acoustic Track, solvers analyzed audio recordings of people speaking, parsing what was said and how it was said to track stages and types of dementia against normative healthy aging.
See the results announcements for each phase for more information on the winning submissions and the participants who developed them. All of the prize-winning report submissions from this competition are available on GitHub for anyone to use and learn from.
- Meet the winners of Phase 1, focused on data discovery and curation
- Meet the winners of Phase 2, who developed top models and wrote the best reports on early prediction from social determinants of health and voice recordings
- Meet the winners of Phase 3, who created the strongest methods and insights for real-world AD/ADRD prediction
- Dive into the winning reports from all phases on GitHub
Phase 2 and 3 solutions were developed using data from the MHAS (Mexican Health and Aging Study) and the DementiaBank.
MHAS is partly sponsored by the National Institutes of Health/National Institute on Aging (grant number NIH R01AG018016) in the United States and the Instituto Nacional de Estadística y Geografía (INEGI) in Mexico. Data files and documentation are public use and available at www.MHASweb.org.
Lanzi, A. M., Saylor, A. K., Fromm, D., Liu, H., MacWhinney, B., & Cohen, M. (2023). DementiaBank: Theoretical rationale, protocol, and illustrative analyses. American Journal of Speech-Language Pathology. doi.org/10.1044/2022_AJSLP-22-00281
Phase 1
Data for Early Prediction (Phase 1)
Find, curate, or contribute data to help the NIH National Institute of Aging create representative and open datasets that can be used for the early prediction of Alzheimer's disease and related dementias. #health
Phase 2
Model Arena - Acoustic Track
Advance algorithms and analytic approaches for early prediction of Alzheimer's disease and related dementias based on voice recordings, with an emphasis on explainability of predictions. [Model Arena - Acoustic Track] #health
Model Arena - Social Determinants Track
Advance algorithms and analytic approaches for early prediction of Alzheimer's disease and related dementias based on social determinants of health, with an emphasis on explainability of predictions. [Model Arena - Social Determinants Track] #health
Report Arena
Advance algorithms and analytic approaches for early prediction of Alzheimer's disease and related dementias, with an emphasis on explainability of predictions. [Report Arena] #health
Phase 3
Proof of Principle Demonstration (Phase 3)
Advance algorithms and analytic approaches for early prediction of Alzheimer's disease and related dementias, with an emphasis on explainability of predictions. [Phase 3] #health