The way we walk (gait) can tell us about our health, risk of falls and brain function, and gait problems are focal in Parkinson’s disease. Gait can be measured using wearable technology (e.g. sensors) and thanks to machine learning models we can classify and identify people with Parkinson’s disease.
Using a wearable sensor placed on the lower back, data was collected from 81 participants with PD and 61 healthy older participants (controls) in our laboratory. A variety of commonly used gait characteristics were quantified (spatiotemporal (e.g. step velocity/ gait speed, etc.) and signal-based (e.g. root mean square values, etc.)), and machine learning models were used to identify the best characteristics which could discriminate between people with Parkinson’s disease and controls. Six partial least square discriminant analysis models were used. These were trained on a subset of 210 gait characteristics that were measured.
The most influential characteristics in the classification models were related to: root mean square values, power spectral density, step velocity and step length, gait regularity and age. This study highlights the importance of signal-based gait characteristics in the development of tools to help classify Parkinson’s disease in the early stages of the disease.
Read more here.