On the Use of Machine Learning Techniques and Non-Invasive Indicators for Classifying and Predicting Cardiac...

Area of Science:

• Cardiology
• Biomedical Engineering
• Data Science

Background:

• Ischemic heart disease diagnosis often relies on invasive or extensive clinical testing.
• Machine learning offers potential for streamlining cardiovascular disease diagnosis.
• Resource-efficient diagnostic tools are crucial for widespread clinical applicability.

Purpose of the Study:

• To enhance ischemic heart disease classification and prediction using machine learning.
• To introduce and evaluate novel, non-invasive Campello de Souza features for data collection.
• To assess the clinical applicability and resource efficiency of these new features.

Main Methods:

• Utilized a comprehensive dataset of heart disease cases from a machine learning repository.
• Introduced novel non-invasive Campello de Souza features requiring only a tensiometer and clock.
• Applied and compared various machine learning algorithms for binary heart disease classification.

Main Results:

• Campello de Souza features, including mean arterial pressure, pulsatile blood pressure index, and resistance-compliance indicator, significantly improved classification accuracy.
• Machine learning algorithms demonstrated streamlined diagnostics, reduced errors, and decreased dependency on extensive testing.
• Logistic regression achieved the highest average accuracy when employing these novel features.

Conclusions:

• Novel non-invasive indicators substantially aid in heart disease classification and prediction.
• Machine learning, augmented by Campello de Souza features, offers a resource-efficient approach to cardiovascular disease diagnosis.
• Integration with comprehensive patient evaluation and medical expertise is recommended for optimal diagnostic frameworks.