Exploration of Machine Learning to Identify Community Dwelling Older Adults with Balance Dysfunction Using Short...

Area of Science:

• Gerontology
• Biomedical Engineering
• Machine Learning

Background:

• Fall-related injuries are a significant health concern for older adults.
• Balance dysfunction is a key risk factor for falls in this population.
• Early identification of individuals at risk is crucial for fall prevention.

Purpose of the Study:

• To assess the feasibility of using wearable sensors and machine learning to identify balance dysfunction in community-dwelling older adults.
• To develop an automated and objective method for detecting early signs of impaired balance during walking.

Main Methods:

• 21 community-dwelling older women participated in the study.
• Data collected during normal walking on an instrumented treadmill and a motor control test (MCT).
• Supervised machine learning (Gradient Boosting Machine) applied to accelerometer data from knee and hip sensors.

Main Results:

• The Gradient Boosting Machine algorithm achieved 91.5% average cross-validation accuracy.
• The model demonstrated a high discriminative ability with an Area Under the Curve (AUC) of 0.97.
• Classification of low and high balance function was achieved using 60-second walking data.

Conclusions:

• Wearable sensors combined with machine learning offer a feasible approach for early diagnosis of balance dysfunction in older adults.
• This technology can aid in identifying individuals at higher risk for falls, enabling timely interventions.
• Early detection and intervention can potentially reduce fall incidence, improve quality of life, and lower healthcare expenditures.