Development and validation of a deep learning model for severe mitral stenosis detection from chest X-rays

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Cardiology

Background:

• Chest X-rays (CXRs) are essential diagnostic tools, but diagnosing mitral stenosis (MS) solely from CXR can be challenging.
• Accurate and timely diagnosis of MS is crucial for patient management.

Purpose of the Study:

• To develop and evaluate a deep learning-based artificial intelligence (AI) model for detecting mitral stenosis (MS) using chest X-ray (CXR) images.
• To assess the diagnostic performance of the AI model in identifying MS.

Main Methods:

• A retrospective analysis of 515 posteroanterior CXR images (285 MS patients, 230 controls) was conducted.
• A deep learning AI model was trained, validated, and tested using a 7:2:1 data split.
• Performance metrics included Area Under the Receiver Operating Characteristic Curve (AUC), precision, recall, F1-score, and accuracy. Saliency maps were used for visualization.

Main Results:

• The AI model achieved a high AUC of 0.99 on both validation and test datasets.
• Excellent performance was observed, with precision, recall, and F1-scores ranging from 0.94 to 0.96 across datasets.
• Saliency maps confirmed that the model focused on clinically relevant radiographic features of MS.

Conclusions:

• The developed deep learning AI model demonstrates significant potential for accurate MS detection from CXRs.
• This AI approach could serve as a valuable and accessible screening tool for MS, particularly in resource-limited settings.
• Further validation in diverse clinical settings is warranted.