Automatic measurement of fall risk indicators in timed up and go test

ABSTRACT

Fall risk assessment is usually conducted in specialized centers using clinical tests. Most of the time, these tests are performed only after the occurrence of health problems potentially affecting gait and posture stability. Our aim is to define fall risk indicators that could routinely be used at home to automatically monitor the evolution of fall risk over time. We used the standard Timed Up and Go (T.U.G.) test to classify 43 individuals into two classes of fall risk, namely high- vs low- risk. Several parameters related to the gait pattern and the sitting position included in the T.U.G. test were automatically extracted using an ambient sensor (Microsoft Kinect sensor). We were able to correctly classify all individuals using machine learning on the combination of two parameters among gait speed, step length and speed to sit down. Coupled to an ambient sensor installed at home to monitor the relevant parameters in daily activities, these algorithms could therefore be used to assess the evolution of fall risk, thereby improving fall prevention.

KEYWORDS:

• Fall prevention
• timed up and go test
• machine learning
• depth camera
• elderly people

Acknowledgments

The authors would like to thank the “Unité d’accueil temporaire et d’orientation” team in Fribourg, Switzerland for their help and support. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Declaration of interest

The authors in this study declare that there is no conflict of interest.