Effective cardiac disease classification using FS-XGB and GWO approach

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Cardiovascular diseases (CVDs) are a major global health concern, necessitating early detection.
• Machine learning (ML) algorithms analyzing electrocardiogram (ECG) signals show promise for CVD prediction.
• Effective ML models require significant feature extraction and selection from ECG data to enhance performance and reduce overfitting.

Purpose of the Study:

• To develop an efficient ML model for identifying five types of CVDs using ECG features.
• To employ Grey Wolf Optimization (GWO) for selecting a reduced, optimal feature set from ECG signals.
• To evaluate a novel feature-specific extreme gradient boosting (FS-XGB) classifier against other ML methods.

Main Methods:

• Extraction of pertinent features from ECG signals.
• Application of Grey Wolf Optimization (GWO) for feature selection, reducing dimensionality.
• Development and implementation of a feature-specific extreme gradient boosting (FS-XGB) classifier.
• Comparative analysis of FS-XGB against Gradient Boosting, AdaBoost, Naïve Bayes, and SVM.

Main Results:

• The proposed FS-XGB model achieved a maximum classification accuracy of 98.8% using only seven optimal features.
• Exceptional performance metrics were recorded: 100% precision, 99.8% recall, 100% F1-score, and 98.8% AUC.
• The methodology significantly outperformed existing approaches in terms of feature reduction and predictive accuracy.

Conclusions:

• The GWO-based feature selection combined with FS-XGB offers a highly effective and efficient method for CVD detection from ECGs.
• This approach demonstrates the potential for improved diagnostic tools in cardiology through advanced ML techniques.
• The study highlights the importance of optimized feature selection for robust and accurate ML-based medical diagnoses.