Respiration Disorders Classification With Informative Features for m-Health Applications

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Respiratory Medicine

Background:

• Respiratory disorders are common and linked to severe health issues.
• Current diagnostic methods are invasive and unsuitable for mobile health.
• There is a need for convenient, low-cost diagnostic tools.

Purpose of the Study:

• To develop an automatic, low-cost diagnostic system for respiratory disorders using wearable sensors.
• To utilize microelectromechanical system (MEMS) motion sensors to monitor chest wall expansion.
• To classify various pathological breathing patterns and distinguish healthy individuals from patients.

Main Methods:

• Employing wearable MEMS motion sensors to measure anterior-posterior chest wall diameter changes.
• Extracting novel respiratory features for breathing disorder classification.
• Evaluating six well-known classifiers (including Support Vector Machine and Decision Tree Bagging) on eight pathological breathing patterns.
• Assessing the impact of sensor number, placement, and feature selection on classification performance.

Main Results:

• Achieved high accuracy rates of 97.50% (Support Vector Machine) and 97.37% (Decision Tree Bagging) for classifying eight breathing patterns.
• Demonstrated superior performance with Decision Tree Bagging (accuracy >98%) in binary classification (healthy vs. patients).
• Evaluated classification performance using accuracy, sensitivity, specificity, F1-score, and Mathew correlation coefficient.

Conclusions:

• Wearable MEMS sensor technology provides a viable, non-obtrusive method for respiratory disorder diagnosis.
• The proposed feature extraction and classification approach shows high efficacy for real-time m-health applications.
• This system offers a promising, accurate, and cost-effective solution for remote patient monitoring and diagnosis.