Automatic Diagnosis of Left Valvular Heart Disease Based on Artificial Intelligence Stethoscope

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence in Medicine

Background:

• Valvular heart disease (VHD) is often underdiagnosed, leading to severe complications.
• Early detection and screening are crucial for effective VHD management.

Purpose of the Study:

• To develop an artificial intelligence (AI)-based stethoscope model.
• The model aims to detect specific left-sided VHD conditions: aortic stenosis, aortic regurgitation, mitral stenosis, and mitral regurgitation.

Main Methods:

• A machine learning algorithm was developed using heart sound recordings from an electronic stethoscope.
• The algorithm was trained on a derivation group and validated on a separate testing group.
• Echocardiography served as the gold standard for diagnosis; model performance was evaluated using the area under the receiver-operating characteristic (AU-ROC) curve.

Main Results:

• The AI model demonstrated diagnostic performance across various left-sided VHDs, with AU-ROC values ranging from 0.6426 (mitral stenosis) to 0.7906 (mitral regurgitation).
• For overall left-sided VHD detection in the training group, the model achieved an AU-ROC of 0.8541, with 83.07% sensitivity and 78.26% specificity.
• In the testing group, the AI model achieved 70.00% sensitivity, 73.68% specificity, and an AU-ROC of 0.7554 for left-sided VHD detection.

Conclusions:

• The AI-based stethoscope demonstrates accurate diagnostic capabilities for left-sided VHD.
• This technology has the potential to make routine VHD screening more practical and accessible.