Deep learning automatically distinguishes myocarditis patients from normal subjects based on MRI.

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Myocarditis is myocardial tissue inflammation, posing risks for heart failure and arrhythmias.
• Cardiovascular magnetic resonance imaging (CMRI) is crucial for diagnosing cardiac conditions.
• Accurate CMRI interpretation is vital for timely myocarditis management.

Purpose of the Study:

• To identify optimal CMRI views for differentiating myocarditis from normal cases.
• To leverage deep learning (DL) for automated myocarditis detection in CMRI.
• To enhance diagnostic accuracy in cardiovascular imaging through AI.

Main Methods:

• A deep learning framework was developed for automated myocarditis detection.
• Analysis included single-frame and multi-frame CMRI data from 269 participants (231 myocarditis, 38 controls).
• Various CMRI views and acquisition types were evaluated for diagnostic performance.

Main Results:

• The DL model achieved a weighted accuracy of 96.9% in detecting myocarditis.
• The late gadolinium enhancement (LGE) 2-chamber view yielded the highest diagnostic accuracy.
• Automated detection demonstrated significant potential in distinguishing myocarditis from normal cases.

Conclusions:

• Deep learning effectively distinguishes myocarditis from normal cases using CMRI.
• Specific CMRI views, particularly LGE 2-chamber, are optimal for DL-based myocarditis detection.
• This AI-driven approach shows promise for improving myocarditis diagnosis and patient outcomes.