Deep Convolutional Neural Network-Based Hemiplegic Gait Detection Using an Inertial Sensor Located Freely in a Pocket

Area of Science:

• Biomedical Engineering
• Machine Learning in Healthcare
• Gait Analysis

Background:

• Traditional gait analysis often requires fixed sensor placement, causing discomfort for daily use.
• Smartphones with built-in inertial sensors offer a convenient alternative for continuous monitoring.

Purpose of the Study:

• To develop and validate a machine learning model for distinguishing hemiplegic gait from normal walking using smartphone inertial sensor data.
• To assess the feasibility of using freely located sensors without pre-processing.

Main Methods:

• Utilized a convolutional neural network (CNN) model trained on acceleration and angular velocity data from a six-axis inertial sensor.
• Employed Bayesian optimization for hyperparameter tuning of the CNN model.
• Conducted a clinical trial with 42 participants, including 21 hemiplegic patients.

Main Results:

• The optimized CNN model achieved an accuracy of 0.78, precision of 0.80, and recall of 0.80.
• The model demonstrated strong performance with an area under the ROC curve of 0.80 and an area under the precision-recall curve of 0.84.
• Distinguishing hemiplegic gait was possible using signals from a pocket-worn sensor without pre-processing.

Conclusions:

• A CNN model applied to pocket-worn inertial sensor data can effectively differentiate hemiplegic gait.
• This approach provides a non-invasive, comfortable, and efficient method for gait analysis in daily life.
• The findings support the potential of using readily available smartphone technology for clinical gait assessment.