A new open-source machine learning tool called ActiTect can identify a strong early warning sign of Parkinson's disease and related disorders just by analyzing wrist movement data during sleep. Researchers from multiple European institutions trained the algorithm to detect isolated rapid eye movement sleep behavior disorder (iRBD), a condition where people physically act out their dreams, often years before neurological diseases appear.
The Research
The team built and tested ActiTect on actigraphy recordings from 279 individuals across five different datasets. In their initial development cohort of 78 people, the model achieved an AUROC of 0.95 (where 1.0 is perfect). When tested on a blinded local set of 31 individuals, performance remained strong at 0.86 AUROC. Two completely independent external cohorts showed AUROCs of 0.84 (n=113) and 0.94 (n=57). To simulate real-world deployment, the researchers used a leave-one-dataset-out cross-validation approach, which yielded a consistent AUROC range of 0.84 to 0.89 across all datasets. This demonstrates that ActiTect can generalize across different devices and recording settings without losing accuracy.
Why It Matters
Most people with iRBD go undiagnosed until significant neurodegeneration has already occurred. Actigraphy—using a simple wrist-worn device similar to a fitness tracker—is cheap and scalable. By automating the analysis with a tool like ActiTect, large populations could be screened non-invasively. The study's senior authors from University Hospital Cologne and Aarhus University highlight that the pipeline's automated sleep-wake detection and robust preprocessing make it viable for home use. For the general public, this means earlier detection of conditions like Parkinson's disease, dementia with Lewy bodies, and multiple system atrophy, potentially opening a window for preventive interventions.
What You Can Do
You can help advance this research by participating in studies that use wearable devices to track sleep and movement. If you have a smartwatch or fitness tracker, pay attention to sleep patterns and discuss any concerns with your doctor. The ActiTect code is freely available online, so technically inclined readers can download and validate the tool on their own data.
Source: arXiv q-bio.NC
Curious about your own brain? Take our free adaptive IQ test or try 306 brain training levels.
