Quick Buzz Feed
KAIST, Sungkyunkwan University, and Korea University Anam Hospital collaborated to analyze real-world lifelog data from 1,224 older adults. Their AI technology achieved 96.5% accuracy in identifying the imminent diagnostic risk of cerebrovascular disease. This breakthrough is expected to transform healthcare from a hospital-centered, reactive model to one focused on predictive and preventive care, by detecting subtle changes in daily life patterns at home.
Cerebrovascular disease often leads to severe complications if not treated promptly, yet it typically presents no symptoms in its early stages, making detection challenging. Researchers at KAIST have developed an artificial intelligence (AI) technology designed to overcome this by analyzing real-life daily activity and environmental data from the homes of older adults. This AI identifies digital behavioral markers indicative of cerebrovascular disease risk, based on subtle, hard-to-notice changes in a person's routine and living environment.
The groundbreaking study was a collaborative effort involving KAIST’s Department of Civil and Environmental Engineering, Sungkyunkwan University’s School of Electronic and Electrical Engineering, and Korea University Anam Hospital’s Department of Neurology. The research leveraged extensive lifelog data collected from 1,224 older adults residing in real residential environments by LivOn Care Co., Ltd. The team meticulously analyzed a total of 13,362 two-week lifelog samples, demonstrating the significant potential to detect early warning signs through nuanced shifts in daily life, moving beyond the conventional approach of diagnosing only after the disease has progressed and symptoms have manifested.
The developed AI technology identifies various cerebrovascular disease risk stages by comprehensively analyzing several data points: daily activity patterns, sleep cycles, circadian rhythms, indoor environmental conditions, as well as the individual's age and existing chronic disease information. A significant achievement of this framework is its ability to assess the proximity of a cerebrovascular disease diagnosis by tracking evolving lifestyle patterns. When comparing lifelog data from within four weeks before a diagnosis (the 'imminent diagnostic risk period') against data from 12 weeks before (the 'non-imminent period'), the AI successfully differentiated between these two phases with an impressive accuracy of 96.53%. This suggests that subtle, continuous monitoring could provide critical insights into escalating health risks even before a clinical visit.
A crucial aspect of this study is the application of explainable AI, which not only determines the existence of a risk but also elucidates the specific lifestyle patterns and environmental factors underpinning its judgments. The analysis revealed that older adults in the prodromal phase of cerebrovascular disease frequently exhibited continuous activity between 10 p.m. and 2 a.m., a period typically reserved for sleep preparation. This indicated irregular daily rhythms, such as delayed sleep onset and a reduced distinction between daytime and nighttime activity, were strongly linked to early cerebrovascular disease signals. Furthermore, as individuals approached a diagnosis, there was a noticeable decrease in continuous evening activity (6 p.m. to 10 p.m.) and a corresponding increase in inactive time. Low indoor humidity, indicating a dry indoor environment, also emerged as a significant environmental factor contributing to an imminent diagnostic risk.
The research team anticipates that this advanced AI technology will serve as an invaluable digital healthcare tool. It offers an objective method to monitor the health status of older adults who might struggle to articulate their own conditions clearly. Moreover, it is expected to provide critical early warning indicators for medical professionals and caregivers, enabling proactive intervention. Professor Lisa Lim emphasized that the AI's role is not to replace hospital diagnoses but to act as a bridge, detecting early risk signals in the home environment and guiding patients to timely medical attention. This shift is poised to contribute to a healthcare paradigm that prioritizes prevention and early intervention over reactive treatment.
The research team explicitly clarified that while promising, this study does not claim to predict the exact onset of cerebrovascular disease or serve as a substitute for clinical diagnosis. Instead, it is designed as a supportive technology aimed at facilitating prevention and early medical consultation. For its successful integration into actual clinical practice, further prospective validation in larger and more diverse patient groups will be essential to confirm its reliability and generalizability.
This significant study, with Dr. Jeongyeop Baek from KAIST as the first author, was published on June 2 in *npj Digital Medicine*. This journal, part of the prestigious Nature Portfolio, is a leading international publication in digital healthcare, boasting an impact factor of 15.1 and ranking in the top 0.3% of JCR journals. The research paper is titled 'AI home monitoring for behavioral markers of cerebrovascular disease' and is available via DOI: 10.1038/s41746-026-02836-7. The work received support from the National Research Foundation (NRF) grant funded by the Korea government (Ministry of Science and ICT).