Automatic and Efficient Fall Risk Assessment Based on Machine Learning.

Area of Science:

• Gerontology
• Biomedical Engineering
• Machine Learning

Background:

• Fall risk assessment is crucial for elderly care and prevention programs.
• Traditional methods like the Berg Balance Scale (BBS) can be time-consuming.
• Efficient screening tools are needed to identify individuals at risk of falls.

Purpose of the Study:

• To develop an automated, non-invasive system for fall risk assessment in the elderly.
• To create a reduced version of the BBS (Efficient-BBS or E-BBS) that maintains assessment accuracy.
• To validate the E-BBS system through a pilot study in a hospital setting.

Main Methods:

• Utilized a multi-depth camera human motion tracking system to capture patient performance on BBS tasks.
• Extracted spatio-temporal features from motion data.
• Trained machine learning classifiers to predict BBS scores and developed fall risk predictors for the E-BBS.

Main Results:

• The E-BBS system reduced the number of required BBS tasks by approximately 50%.
• The automated system achieved 97% accuracy in fall risk assessment.
• Pilot study demonstrated the feasibility and effectiveness of the E-BBS in a clinical setting.

Conclusions:

• The automated E-BBS offers an efficient and accurate method for fall risk screening in the elderly.
• This technology can significantly reduce the time and resources required for fall risk evaluations.
• The developed machine learning approach is adaptable for other medical assessments.