A comparison of pressure mat and force plate parameters for classifying elderly fallers

The purpose of this study was to compare elderly faller classifications using pressure mat and force parameters as input to machine learning algorithms. Balance and gait were assessed using pressure mat and force plate data in 59 non-fallers (NF) and 41 fallers (F). Biomechanical parameters included 3D ground reaction force-time data (GRF), centre of pressure displacement/velocity data (COP), and data unique to the pressure mats (PM), including peak contact pressure, mean pressure, and force/pressure time integrals. For both standing balance and gait, these biomechanical parameters served as input to a discriminant analysis machine learning algorithm to build a model and classify fallers. The model resulted in high classification accuracies: 80.65% for pressure mat parameters and 84.95% for the force plate parameters. Combining parameters from walking and standing resulted in the most accurate classifications. Results demonstrated the feasibility of using portable pressure mats in clinical settings to identify fallers.