AI Foundation Models for Wearable Movement Data in Mental Health Research
Author Information
Author(s): Ruan Franklin Y., Zhang Aiwei, Oh Jenny Y., Jin SouYoung, Jacobson Nicholas C.
Primary Institution: Cornell University
Hypothesis
Can pretrained foundation models improve the analysis of wearable movement data for mental health research?
Conclusion
The Pretrained Actigraphy Transformer (PAT) outperforms existing methods in predicting mental health outcomes using wearable movement data.
Supporting Evidence
- PAT is the first open source foundation model designed for time-series wearable movement data.
- The model achieves state-of-the-art performance in several mental health prediction tasks.
- PAT is lightweight and easily interpretable, making it suitable for mental health research.
Takeaway
This study created a new tool that helps researchers understand how movement data from wearables can predict mental health.
Methodology
The study developed the Pretrained Actigraphy Transformer (PAT) using transformer architectures and pretraining on a large dataset of wearable movement data.
Participant Demographics
Data was collected from a national U.S. sample.
Want to read the original?
Access the complete publication on the publisher's website