Smart sensors could flag stroke risk in older people living alone, study finds

 FYI.

Smart sensors could flag stroke risk in older people living alone, study finds

Smart home sensors may help flag stroke risk in older people living alone by detecting changes in daily behaviour, a South Korean study suggests.

The AI-powered system used contactless Internet of Things sensors to identify behavioural patterns linked to a higher risk of stroke.

Researchers said monitoring these changes could support earlier detection among older adults.

Dr Cho Kyung-hee, professor of neurology at Korea University Anam Hospital and co-lead investigator, said: “Early intervention is crucial for the prognosis of cerebrovascular disease, but it is easy to miss subtle changes in elderly patients.”

Researchers from Korea University Anam Hospital, the Korea Advanced Institute of Science and Technology and Sungkyunkwan University collected smart home data from 1,224 people in South Korea.

All participants were aged 65 or older, lived alone and could walk independently.

They were divided into three groups: people with no history of cerebrovascular disease, those previously diagnosed with the condition and a prodromal group later found to be at an early stage of the disease.

Cerebrovascular disease affects the blood vessels supplying the brain and includes conditions that can lead to stroke.

Prodromal refers to an early stage in which subtle changes may appear before a condition is diagnosed.

Researchers analysed more than 13,000 data points collected over 14 days using motion, door, temperature and humidity sensors.

The data covered physical activity, periods of inactivity, sleep patterns and indoor conditions.

Six AI models were trained to recognise changes and patterns in the sensor data.

The best-performing model, TabNet, recorded an AUPRC of 85 per cent when identifying participants in the prodromal group.

AUPRC measures how well a model identifies people in a target group while limiting false alerts.

The model recorded an AUROC of 91 per cent when distinguishing participants with a previous diagnosis from those with no history of cerebrovascular disease.

AUROC measures how effectively a system separates two groups, with a higher score indicating better performance.

In retrospective testing, the model identified home activity patterns recorded during the four weeks before a stroke diagnosis with 95.12 per cent sensitivity, 96.97 per cent specificity and 96.53 per cent accuracy.

Sensitivity measures how well a test identifies people with a condition, while specificity measures how accurately it rules out those without it.

More frequent periods of sustained movement before sleep, less time spent inactive and a later sleep onset were key markers of the prodromal group.

Among participants with a previous diagnosis, the main indicators included more continuous activity during the night and more frequent sleep interruptions.

Greater evening inactivity, fewer periods of sustained activity and low or high indoor humidity were also linked to a higher likelihood of diagnosis within four weeks.

The researchers said AI-powered contactless monitoring could eventually complement clinical assessments by helping to identify possible warning signs earlier.

They said the approach could be particularly useful for older adults living alone who may not immediately recognise changes in their health.

Hospitals across the Asia-Pacific region have also introduced AI-powered stroke imaging platforms in an effort to reduce treatment delays.

Last year, Apollo Hospitals in Chennai began implementing an AI-driven imaging system that reportedly reduced the time from scanning to image interpretation from about 30 minutes to seven minutes.