Detecting Congestive Heart Failure by Extracting Multimodal Features and Employing Machine Learning Techniques

Area of Science:

• Cardiology and Biomedical Engineering
• Computational Biology and Machine Learning

Background:

• Heart rate variability (HRV) reflects cardiac adaptability to stimuli.
• Reduced HRV is a predictor of negative cardiovascular outcomes.
• Linear HRV measures are limited in analyzing complex cardiovascular dynamics.

Purpose of the Study:

• To develop an automated system for analyzing HRV signals.
• To extract multimodal features capturing temporal, spectral, and complex dynamics.
• To evaluate machine learning techniques for detecting congestive heart failure.

Main Methods:

• Utilized multimodal features (temporal, spectral, complex dynamics) from HRV signals.
• Employed machine learning classifiers: Support Vector Machine (SVM), Decision Tree (DT), k-Nearest Neighbor (KNN), and ensemble methods.
• Evaluated performance using specificity, sensitivity, PPV, NPV, and AUC.

Main Results:

• The automated system achieved high detection performance for congestive heart failure.
• SVM linear kernel yielded the highest performance (93.1% total accuracy, 0.97 AUC).
• Ensemble subspace discriminant and SVM medium Gaussian kernel also showed strong results (91.4% and 90.5% accuracy, respectively).

Conclusions:

• The proposed automated HRV analysis system is effective for detecting congestive heart failure.
• The approach offers a computationally efficient tool for clinical application.
• Advanced machine learning techniques enhance the analysis of complex HRV dynamics.