A team at Korea University integrates Fitbit-derived activity and heart-rate metrics with nightly app entries using cosinor-based circadian features to train random forest and XGBoost classifiers, distinguishing moderate and severe RLS symptom groups with AUCs up to 0.86.
Key points
Integration of 85 circadian-based features from Fitbit Inspire wearables and the SOMDAY smartphone app
Random Forest model achieved AUC 0.86 for moderate RLS prediction; XGBoost reached AUC 0.70 for severe RLS prediction
SHAP analysis highlighted M10 step counts, relative amplitude, and stress level as primary predictive features
Why it matters:
Objective digital phenotyping and ML screening could revolutionize early detection and personalized management of RLS, reducing diagnostic delays due to subjective reporting.
Q&A
What is digital phenotyping?
How do circadian features improve prediction?
What role does SHAP analysis play?
Read full article
Academy
Circadian Rhythms and Aging
Circadian rhythms are 24-hour biological cycles that regulate sleep, hormone release, metabolism, and cellular repair. In aging populations, these rhythms often weaken, leading to sleep disturbances, reduced metabolic efficiency, and impaired tissue recovery. Maintaining robust circadian patterns supports healthy aging by aligning physiological processes with environmental cues.
Key elements of circadian biology:
- Suprachiasmatic nucleus (SCN): The brain’s master clock, located in the hypothalamus, synchronizes peripheral clocks throughout the body.
- Light exposure: Daytime sunlight signals reset the SCN, while nighttime darkness promotes melatonin secretion and restorative sleep.
- Melatonin: A hormone released by the pineal gland; its nighttime surge facilitates sleep onset and supports antioxidant defenses.
- Peripheral clocks: Found in organs such as the liver and muscle, these regulate tissue-specific functions like detoxification and glucose metabolism.
Age-related changes: With age, SCN neuron numbers decline and melatonin production decreases. Disrupted sleep-wake cycles impair memory consolidation and increase inflammation, contributing to cognitive decline and metabolic disorders.
Digital Phenotyping in Health and Longevity
Digital phenotyping is the real-time collection and analysis of data from smartphones, wearables, and apps to measure health-related behaviors and physiology. For longevity science, digital phenotyping offers continuous monitoring of activity levels, sleep quality, and heart rate variability—metrics linked to aging trajectories.
Applications to healthy aging:
- Early detection: Subtle changes in sleep patterns or daily activity can signal emerging conditions like frailty or metabolic decline before clinical symptoms appear.
- Personalized interventions: Data-driven insights enable tailored lifestyle recommendations—such as optimized exercise timing or light therapy—to reinforce circadian alignment and boost resilience.
- Longitudinal tracking: Continuous monitoring reveals individual responses to dietary or pharmacological interventions, guiding adaptive regimens for maximal healthspan benefits.
By integrating digital phenotyping with circadian biology, longevity researchers can design interventions that preserve rhythm strength, enhance cellular repair, and ultimately extend healthspan. This non-invasive approach democratizes aging research and empowers individuals to track their biological age using everyday devices.
