Finding signatures of Parkinson’s disease in patients’ gait

For Parkinson’s disease (PD) patients, early detection is crucial for quality-of-life outcomes, and changes in a person’s gait are one of the first signs of the disease. Researchers and doctors frequently evaluate PD patients’ stride timing, balance control, and rhythm, but the field lacks a standardized approach to using these measurements.

Using a method called recurrence triangle (RT) analysis, Hasan et al. showed subtle patterns hidden within gait signals can reliably distinguish patients with PD from healthy controls. In tests, their method worked with 100% accuracy.

“Think of watching someone walk and noting every moment when their stride looks similar to a previous moment,” said author Md. Mehedi Hasan.

Placing all these moments on a grid creates a pattern that reflects the person’s rhythm. RT analysis searches for tiny triangles of matching points within this pattern — small repetitions in movement.

“People with healthy motor control create one type of triangle pattern; people with Parkinson’s create a noticeably different one, because PD affects timing, symmetry, and consistency of movement,” said Hasan. “The frequency of these triangles becomes a stable ‘signature’ of the person’s gait, and that signature is what we use for classification.”

With devices collecting velocity and acceleration data at their hips and both ankles, the researchers compared 25 people with PD walking 15 meters — about 50 feet — to 25 healthy controls. Each data type had over 90% accuracy in predicting PD using RT analysis on its own, but using a combination of all the data types — velocity and acceleration at all three locations — yielded a perfect prediction performance.

The method is computationally simple and easy to interpret, enabling real-time diagnoses and fostering clinical trust. The researchers plan to extend it to include comparisons to ALS, as well as other mobility outcomes of PD.

Source: “Small sample learning classifies Parkinson’s disease patients based on their walking behavior,” by Md. Mehedi Hasan, Takaaki Hattori, and Yoshito Hirata, Chaos (2025). The article can be accessed at https://doi.org/10.1063/5.0288442 .