Identification of optimal classifier and sensor placement for fall risk classification using IMU-based gait data.

Area of Science:

• Biomedical Engineering
• Gerontology
• Rehabilitation Science

Background:

• Fall risk assessment in older adults is crucial for prevention.
• Gait analysis using inertial measurement units (IMUs) shows promise for classifying fall risk.
• Optimal IMU sensor placement and classification algorithms for fall risk remain underexplored.

Purpose of the Study:

• To determine the optimal placement strategy for IMU sensors to maximize the accuracy of fall risk classification during gait.
• To compare the performance of various machine learning algorithms for fall risk classification using IMU data.

Main Methods:

• Ninety-three older adults were assessed for fall risk (low vs. high).
• Kinematic data were collected using 10 IMUs placed on various body segments during a 10m walk.
• Features extracted included mean and variance of linear acceleration and angular velocity.
• Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), K-Nearest Neighbors (KNN), XGBoost, and LightGBM classifiers were evaluated.

Main Results:

• Sensor placement significantly impacted classification performance (p < 0.001).
• The lower leg emerged as the optimal sensor placement, achieving the highest accuracy.
• The SVM classifier using a left lower leg sensor achieved 90.1% accuracy, 95.7% sensitivity, and 84.1% specificity.
• No significant effects were found for the classification algorithm or its interaction with sensor placement (p > 0.05).

Conclusions:

• IMU sensor placement is a critical factor in fall risk classification accuracy.
• Placing IMU sensors on the lower leg provides the highest accuracy for fall risk assessment.
• These findings offer valuable insights for developing advanced fall prevention technologies for the elderly.