Machine learning-based assessment tool for imbalance and vestibular dysfunction with virtual reality rehabilitation...

Area of Science:

• Neuroscience
• Rehabilitation Medicine
• Human-Computer Interaction

Background:

• Dizziness, often caused by imbalance and vestibular dysfunction, significantly impacts quality of life.
• While balance training offers a non-invasive alternative to surgery or drugs, traditional methods are tedious and lack effective assessment tools.
• Current diagnostic methods for vestibular dysfunction are insufficient for rapid patient severity assessment.

Purpose of the Study:

• To develop and evaluate an interactive virtual reality (VR) game-based rehabilitation program for balance training.
• To introduce a sensor-based system for quantifying balance indices.
• To assess the therapeutic efficacy and diagnostic capability of the VR program and sensor system.

Main Methods:

• An interactive VR game incorporating Cawthorne-Cooksey exercises was developed.
• A sensor-based measuring system was integrated for real-time balance data collection.
• A clinical experiment involving 48 patients and 36 healthy subjects was conducted, with data analyzed using statistical tools and a Support Vector Machine (SVM) classifier.

Main Results:

• Patients demonstrated significant improvement in balance indices after completing the VR training program.
• Quantified balance data clearly distinguished between patients with vestibular dysfunction and healthy individuals.
• The SVM classifier achieved high accuracy in differentiating between patient and control groups.

Conclusions:

• The VR game-based balance training program is effective in improving patient outcomes.
• The sensor-based system and SVM analysis provide a feasible method for rapid assessment of patient severity and diagnosis of vestibular dysfunction.
• This approach offers a more engaging and objective alternative to traditional balance rehabilitation and assessment.