Your Daily Routine Could Reveal Stroke Risk Weeks Before Symptoms

Small changes in how you sleep, move, and spend your evenings at home could signal that a stroke or other cerebrovascular disease is approaching — weeks before any obvious medical symptoms appear. A team of researchers from KAIST, Sungkyunkwan University, and Korea University Anam Hospital has developed an artificial intelligence system that can detect these early warning signs with 96.5% accuracy.

The study, published in the journal npj Digital Medicine, analyzed 13,362 two-week lifelog records collected from 1,224 older adults living in their own homes. The lifelogs captured daily movement patterns, sleep timing, activity levels across different times of day, indoor humidity, and other environmental data, combined with basic health information such as age and chronic conditions.

The AI identified several distinctive patterns that precede a cerebrovascular diagnosis. Older adults entering a pre-diagnostic risk stage showed disrupted circadian rhythms — they tended to remain active between 10 p.m. and 2 a.m., when most people naturally sleep. As the diagnosis became more imminent, activity levels between 6 p.m. and 10 p.m. dropped noticeably, while total inactive time increased. A dry indoor environment with low humidity also emerged as a significant contributing factor.

What makes this approach different from conventional medical screening is that it is entirely passive. The system does not require blood tests, brain scans, or clinical visits. It works by observing everyday behaviour that the person may not even notice is changing. Because the AI is "explainable," it can show doctors exactly which behavioural and environmental factors contributed to its risk assessment, rather than producing a black-box prediction.

Three essential facts to understand: First, the AI distinguished between data collected four weeks before diagnosis and data from twelve weeks earlier with 96.53% accuracy — meaning lifestyle changes become progressively more detectable as the disease approaches. Second, the system does not replace hospital diagnosis but serves as an early warning layer that can prompt medical assessment before major symptoms develop. Third, the technology represents a shift from reactive medicine — treating disease after it has already caused damage — to proactive prevention, where the home itself becomes a continuous health monitoring environment.

The researchers caution that larger prospective studies are needed before the system can be deployed at scale. But the results suggest that the way we live each day may contain health signals that are invisible to the naked eye — and that AI is now learning to read them.