When Professor Lisa Lim started studying how older adults move around their homes, she wasn't just watching where they went — she was listening to whispers of their health.
At KAIST, a university in Daejeon, South Korea, Lim and her team have created an AI tool that learns a person's daily habits — when they sleep, how much they move, even how humid their home is — and looks for tiny shifts that might signal trouble ahead. Their goal is to catch the early warning signs of cerebrovascular disease, the medical term for problems with blood flow to the brain that can lead to stroke, before symptoms become impossible to ignore.
"The key point of this study is not that AI should replace a hospital diagnosis, but that it can first detect risk signals in small lifestyle changes at home and help connect patients to medical care at the right time," Lim said.
The research, published with collaborators from Sungkyunkwan University and Korea University Anam Hospital, analyzed lifelog data from 1,224 older adults. That's 13,362 two-week snapshots of real life in real homes, collected by LivOn Care Co., Ltd. The AI learned to spot patterns — like a person suddenly staying up late more often, or moving less in the evening — and could tell the difference between someone four weeks away from a cerebrovascular disease diagnosis and someone twelve weeks away with 96.53 percent accuracy.
The science behind it is striking. People in the early stages of cerebrovascular disease tended to show frequent activity between 10 p.m. and 2 a.m., when the body normally slows down for sleep. Their days and nights started to blur. As diagnosis loomed closer, they moved less during evening hours between 6 p.m. and 10 p.m., sitting idle more often. Surprisingly, dry indoor air also showed up as a warning sign.
Unlike some AI systems that work like black boxes — giving answers without explanation — this tool can show why it flagged a concern. It tells caregivers and doctors exactly which habits changed and how, making it easier for everyone to understand.
Of course, the researchers are careful not to overpromise. This doesn't predict strokes or replace doctor visits. It's meant to be a helper — a digital canary in the coal mine that nudges families and doctors to pay closer attention before a crisis hits. More testing with larger groups will be needed before hospitals could actually use it.
Still, the hope is real. For older adults who might not notice subtle changes in themselves or struggle to describe how they feel, this technology could serve as an extra set of eyes — one that watches quietly from the corner and says, "Maybe it's time to see a doctor."