Sensor-Based Multifaceted Feature Extraction and Ensemble Elastic Net Approach for Assessing Fall Risk in...

Area of Science:

• Gerontology
• Biomedical Engineering
• Data Science

Background:

• Accurate fall risk identification in community-dwelling older adults is crucial for timely intervention and reducing fall incidents.
• Sensor-based motion analysis of gait and balance is a promising approach for fall risk assessment.
• Existing methods often overlook non-linear signal characteristics like complexity and local stability.

Purpose of the Study:

• To develop a comprehensive feature extraction scheme for fall risk assessment.
• To investigate the utility of non-linear features from sensor data in multi-task scenarios.
• To propose an interpretable and accurate machine learning model for fall risk classification.

Main Methods:

• Extracted multifaceted features (demographic, statistical, non-linear, spatiotemporal, spectral) from accelerometer and gyroscope data.
• Developed an ensemble elastic net (E-E-N) model using bootstrap sampling and weighted aggregation.
• Investigated non-linear features from a dynamic system perspective for the first time in this context.

Main Results:

• The proposed multifaceted feature extraction scheme captured diverse signal characteristics.
• The ensemble elastic net (E-E-N) approach demonstrated superior predictive performance in fall risk classification.
• Validation experiments using real-world data confirmed the model's effectiveness.

Conclusions:

• The developed approach offers a cost-effective tool for accurate fall risk assessment in older adults.
• This method can help alleviate the burden of long-term continuous health monitoring.
• Integrating non-linear features enhances the accuracy of fall risk prediction models.