Screening of Ischemic Heart Disease based on PPG Signals using Machine Learning Techniques

Area of Science:

• Biomedical Engineering
• Cardiology
• Signal Processing

Background:

• Cardiac ailments are rising, necessitating accessible diagnostic tools.
• Current heart disease detection methods are expensive and require specialized equipment.
• There is a critical need for low-cost, easily deployable screening techniques.

Purpose of the Study:

• To explore the potential of finger-tip photoplethysmography (PPG) signals for Ischemic Heart Disease (IHD) detection.
• To develop an accessible and affordable screening method for IHD.
• To evaluate machine learning classifiers for IHD identification using PPG signals.

Main Methods:

• Collected and analyzed time-domain features from finger-tip PPG signals.
• Employed various machine learning algorithms including Decision Trees, Discriminant Analysis, Logistic Regression, Support Vector Machine, KNN, and Boosted Trees for classification.
• Utilized confusion matrix to assess ten performance metrics, including accuracy, sensitivity, and specificity.

Main Results:

• The Boosted Trees classifier achieved high performance metrics: 0.94 accuracy, 0.95 sensitivity, 0.95 specificity, and 0.97 precision.
• Receiver Operating Characteristic (ROC) and Area Under the Curve (AUC) were calculated to validate classification robustness.
• The study demonstrated the effectiveness of PPG signal analysis for IHD patient identification.

Conclusions:

• Finger-tip PPG signal analysis presents a promising, low-cost, and accessible method for IHD screening.
• This technique can significantly benefit individuals in remote and underserved regions.
• Machine learning classifiers, particularly Boosted Trees, show strong potential for reliable IHD detection using PPG data.