From Heart Murmur to Echocardiography - Congenital Heart Defects Diagnostics Using Machine-Learning Algorithms

Area of Science:

• Pediatric Cardiology
• Biomedical Engineering
• Machine Learning

Background:

• Heart murmurs are common in pediatric cardiology but often lack specific characteristics.
• Distinguishing organic murmurs associated with congenital heart defects (CHD) is crucial for timely intervention.

Purpose of the Study:

• To develop and evaluate a machine learning classifier for identifying organic heart murmurs indicative of CHD in pediatric patients.
• To assess the performance of various machine learning algorithms in classifying heart murmurs.

Main Methods:

• Utilized a dataset of 116 children (aged 1-15 years) with heart murmur data collected over three years.
• Performed feature relevance analysis using InfoGain, GainRatio, Relief, and Correlation methods.
• Developed and compared classifiers including Naive Bayes, Logistic Regression, Decision Tree, Random Forest, and Support Vector Machine.

Main Results:

• Achieved high classification accuracy (>90%) for detecting CHD using 16 input parameters.
• Demonstrated the effectiveness of machine learning in differentiating organic murmurs from other types.
• Identified key parameters from basic physical examination for accurate classification.

Conclusions:

• A highly accurate machine learning classifier for CHD detection based on heart murmur characteristics can be developed.
• Such a tool can serve as a valuable diagnostic aid for pediatricians and primary healthcare providers.
• Early and accurate diagnosis of CHD through intelligent systems can significantly improve pediatric cardiac care.