Discrimination of vestibular function based on inertial sensors

Area of Science:

• Biomedical Engineering
• Neurology
• Gait Analysis

Background:

• Vestibular dysfunction commonly causes gait and balance abnormalities.
• Current detection methods are static and fail to capture dynamic patient information.
• A simple, dynamic detection method for vestibular dysfunction is needed.

Purpose of the Study:

• To propose a simple method for detecting vestibular dysfunction using gait signals.
• To evaluate the accuracy of inertial sensor-based gait analysis for diagnosing vestibular dysfunction.

Main Methods:

• Gait data acquired using dynamic gait index (DGI) and inertial sensors.
• Extracted time-domain, frequency-domain, and non-linear features from sensor signals.
• Utilized Relief algorithm for feature selection and SVM/RF classifiers for diagnosis.

Main Results:

• Magnetometer features with SVM achieved 84.79% accuracy.
• Combining features from three sensor signals with RF classifier yielded 86.5% accuracy.
• Demonstrated the effectiveness of inertial sensor data for classifying vestibular dysfunction.

Conclusions:

• Inertial sensor-based detection of vestibular dysfunction is simple, accurate, and easily implementable.
• This method offers a novel approach for clinical diagnosis of vestibular function.
• Dynamic gait analysis provides valuable insights into vestibular health.