Algorithm that analyses smartwatch data detects minute movement changes many years before symptoms appear.
Summary
Smartwatches could spot if someone has Parkinson’s disease many years before they notice obvious symptoms. An artificial-intelligence algorithm, created with 104,000 UK Biobank participants’ activity data, could detect minute subtle slowing-down of movements up to seven years before someone received a Parkinson’s diagnosis.
An artificial-intelligence (AI) algorithm that analyses motion data from a smartwatch can detect Parkinson’s disease years before someone notices symptoms such as tremors or muscle stiffness. This could help people to get diagnosed much earlier and access the support they need – or receive potential future treatments.
We know that if we want to treat Parkinson’s, we need to target people before all this early [brain cell] damage occurs.
Dr Kathrin Schalkamp, Cardiff University, UK
Parkinson’s disease is usually diagnosed only when someone notices movement problems. This can be many years after the condition started to affect the brain. “We know that if we want to treat it, we need to target people before all this early [brain cell] damage occurs,” explained medical researcher Ann-Kathrin Schalkamp in Discover Magazine.
A better predictor for Parkinson’s
Schalkamp was part of a team that trained an AI algorithm on data from nearly 104,000 UK Biobank participants who wore activity-tracking smartwatches for a week between 2013 and 2015. The algorithm detected that people’s movements minutely slowed down up to seven years before they received a diagnosis of Parkinson’s disease. It could also distinguish the condition from the other things that might change someone’s movement such as having osteoarthritis or simply getting older.
The algorithm was better than other methods, including genetic or blood biochemistry tests, at predicting who would develop Parkinson’s disease. There are no drugs yet that can slow down or stop the condition’s progression, but more than 100 are being tested. Smartwatches with AI algorithms could be an important, low-cost way to match people in the very early stages of Parkinson’s disease with clinical trials for these treatments, study leader Cynthia Sandor told the i Paper.
The pros and cons of early diagnosis
Before such a tool becomes available, the study’s results would have to be repeated, Sandor cautions. And the algorithm would need to be adapted for regular consumer smartwatches rather than the medical ones given to UK Biobank participants.
“The research is very interesting,” Parkinson’s researcher José López Barneo told the Science Media Centre. But it’s not clear if people would benefit from an earlier diagnosis, he adds, given the current lack of treatments. Whether people should be told that they have Parkinson’s disease years before symptoms develop “will always remain an individual and personal choice”, study team member Kathryn Peall told BBC News.