AI Drawing Test Shows Promise for Detecting Parkinson Disease

A small study found AI could distinguish people with Parkinson disease from healthy controls using smart-pen drawings, reaching nearly 99% accuracy.

Mind

Details

A simple drawing exercise may offer a new way to study Parkinson disease. In a recent experiment, researchers used data from 66 people, including 31 with Parkinson disease and 35 healthy controls. Participants traced spirals and angular meander shapes with a smart pen that recorded both the drawings and details of hand movement such as pressure, grip, tilt, and acceleration. AI models then looked for patterns in the shapes and the movement signals. The combined system distinguished the Parkinson disease group from the healthy control group with reported accuracy of about 98 percent for the meander drawings and about 98 percent for the spiral drawings. The appeal is that drawing is inexpensive, non-invasive, and easy to repeat, while sensor data can capture movement details that are difficult to judge by eye. However, the results are still preliminary. The dataset was small, and the participants already belonged to known Parkinson disease or healthy groups. That means the study does not establish that the test can predict who will develop Parkinson disease in the future, nor does it show that the method can reliably distinguish Parkinson disease from every other movement disorder. Larger and more diverse clinical studies would be needed before a drawing-based AI system could be considered a dependable screening or diagnostic tool. The research is therefore best viewed as an encouraging demonstration of what detailed handwriting and drawing measurements might eventually contribute to neurological assessment.

More videos

Up nextStudy Maps How Slow Nasal Breathing May Calm Anxiety