Eigen Posture Based Fall Risk Assessment System Using Kinect.

Area of Science:

• Geriatric Medicine
• Biomedical Engineering
• Machine Learning

Background:

• Postural Instability (PI) significantly increases fall risk in the elderly and patients with neurological disorders.
• Conventional fall risk assessments, like the Berg Balance Scale (BBS), require clinical settings and extensive patient participation.
• There is a need for accessible, low-intervention methods to accurately assess fall risk.

Purpose of the Study:

• To develop and validate a machine learning-based system for fall risk assessment.
• To utilize the Single Limb Stance (SLS) exercise for PI detection with minimal human intervention.
• To offer an affordable and unobtrusive solution for classifying fall risk categories.

Main Methods:

• Employing a Kinect sensor to capture spatiotemporal dynamics of skeleton joint positions during SLS.
• Developing a novel posture modeling technique for feature extraction.
• Integrating traditional time-domain and metadata features with novel posture features.

Main Results:

• The system achieved a 75% mean accuracy in predicting fall risk categories.
• The approach was validated on a diverse cohort of 224 subjects, including geriatric and patient populations.
• Demonstrated the efficacy of using SLS exercise data for fall risk assessment.

Conclusions:

• The proposed machine learning system provides an effective and accessible method for fall risk assessment.
• Minimal human intervention and the use of SLS exercise make the system practical for wider application.
• This technology has the potential to improve fall prevention strategies in clinical and home settings.