Machine learning based congestive heart failure detection using feature importance ranking of multimodal features

Area of Science:

• Cardiology
• Biomedical Engineering
• Machine Learning

Background:

• Congestive Heart Failure (CHF) poses a significant global health challenge.
• Accurate and early detection of CHF is crucial for effective patient management.
• Current diagnostic methods can be further enhanced by advanced computational approaches.

Purpose of the Study:

• To rank multimodal features for improved detection of Congestive Heart Failure (CHF) subjects.
• To evaluate the efficacy of machine learning models using selected high-ranking features.
• To compare the performance of different machine learning algorithms in CHF detection.

Main Methods:

• Multimodal features were extracted from Congestive Heart Failure (CHF) and Normal Sinus Rhythm (NSR) subjects.
• Features were ranked and categorized using Empirical Receiver Operating Characteristics (EROC) values.
• Machine learning models including Decision Tree (DT), Naïve Bayes (NB), and various Support Vector Machines (SVM) were employed.

Main Results:

• The highest accuracy (88.79%) and AUC (0.95) were achieved using all multimodal features with SVM Gaussian.
• Using top-ranked features, SVM Gaussian (top 5) achieved 84.48% accuracy and 0.86 AUC.
• Decision Tree and Naïve Bayes with top 9 features yielded 84.48% accuracy and 0.88 AUC.

Conclusions:

• Feature ranking significantly aids in the automatic detection of CHF.
• The proposed approach offers a valuable tool for clinicians to improve diagnostic decisions.
• This method has the potential to reduce CHF-related mortality rates.