CNN-KCL: Automatic myocarditis diagnosis using convolutional neural network combined with k-means clustering

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Imaging

Background:

• Myocarditis, heart wall inflammation, is a difficult diagnosis in cardiology.
• It is a leading cause of sudden death in adults under 40.
• Cardiac MRI (CMR) is a key diagnostic tool but is limited by interpretation variability and technical factors.

Purpose of the Study:

• To introduce a novel deep learning model for diagnosing myocarditis.
• To evaluate the efficacy of the Convolutional Neural Network-Clustering (CNN-KCL) model.

Main Methods:

• Development of a deep learning model named Convolutional Neural Network-Clustering (CNN-KCL).
• Utilized 98,898 images from 47 subjects for training and validation.
• Employed a 10-fold cross-validation technique with 4 clusters.

Main Results:

• The CNN-KCL model achieved a diagnostic accuracy of 97.41% for myocarditis.
• Demonstrated high performance in identifying myocarditis from cardiac images.
• This represents the first application of deep learning for myocarditis diagnosis.

Conclusions:

• The CNN-KCL deep learning model shows significant potential for accurate myocarditis diagnosis.
• AI-driven approaches can overcome limitations of traditional CMR interpretation.
• This research paves the way for advanced diagnostic tools in cardiology.