KAIST Develops AI to Detect Brain Vascular Disease Risk Signals with 96.5% Accuracy

KAIST, in collaboration with researchers from Sogang University, Seoul National University Bundang Hospital, and Korea University Anam Hospital, has developed an AI framework capable of identifying early risk stages for brain vascular diseases and assessing imminent diagnosis conditions. The study utilized LiProg data from 1,224 elderly individuals collected over extended periods in real residential environments. By analyzing 13,362 daily living data samples, the team demonstrated the potential to detect risk signals through subtle changes in daily routines. The AI framework simultaneously analyzes lifestyle patterns, sleep, circadian rhythms, indoor environments, age, and chronic conditions. It successfully distinguished between imminent (within 4 weeks of diagnosis) and non-imminent (more than 12 weeks before diagnosis) phases with 96.53% accuracy. This explainable AI provides actionable insights into lifestyle patterns and environmental factors linked to risk signals, such as irregular sleep schedules and decreased activity during specific hours. The technology is positioned as a digital healthcare tool for objective monitoring and early warning systems, though it is emphasized as a supportive diagnostic aid rather than a replacement for clinical diagnosis.