Automated Phenotyping of Mitral Stenosis Using Deep Learning.

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Imaging

Background:

• Accurate classification of mitral stenosis (MS) is a clinical challenge.
• Echocardiography is a key diagnostic tool for MS.
• Automated detection of MS can improve patient care.

Purpose of the Study:

• To develop an artificial intelligence (AI) framework for automatic detection of clinically significant MS.
• To assess MS severity and differentiate rheumatic etiology using echocardiography.
• To validate the AI framework's performance across multiple healthcare cohorts.

Main Methods:

• Developed EchoNet-MS, an open-source, end-to-end AI framework.
• Utilized video-based convolutional neural networks for MS assessment.
• Trained and validated the model on over 431,000 echocardiography videos from four distinct cohorts.

Main Results:

• EchoNet-MS demonstrated excellent discrimination of severe MS with AUCs ranging from 0.937 to 0.994 across cohorts.
• The AI model achieved high performance in classifying rheumatic vs. non-rheumatic MS (AUC 0.890-0.967).
• The model showed robust generalization across external validation cohorts.

Conclusions:

• EchoNet-MS accurately assesses MS severity and etiology from echocardiograms.
• The AI tool shows strong, generalized performance across diverse patient populations.
• EchoNet-MS has potential as an automated clinical decision support tool for MS detection.