RLMD-PA: A Reinforcement Learning-Based Myocarditis Diagnosis Combined with a Population-Based Algorithm for...

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Imaging

Background:

• Myocarditis, heart muscle inflammation, is increasingly prevalent, particularly with COVID-19.
• Cardiac magnetic resonance (CMR) imaging is crucial for diagnosis but requires expert interpretation.
• Automated systems can improve the efficiency and accuracy of CMR image analysis for myocarditis.

Purpose of the Study:

• To develop and evaluate an automatic computer-aided diagnostic system for myocarditis classification using CMR images.
• To address the challenge of imbalanced datasets in medical image classification.
• To formulate myocarditis classification as a sequential decision-making process.

Main Methods:

• A deep reinforcement learning approach, RLMD-PA, was developed, utilizing a convolutional neural network (CNN) policy.
• The artificial bee colony (ABC) algorithm was employed for initial weight optimization.
• The model was trained on the Z-Alizadeh Sani myocarditis dataset, with a reward function prioritizing minority class identification.

Main Results:

• The RLMD-PA model demonstrated high accuracy in classifying myocarditis from CMR images.
• The system effectively handled imbalanced data, a common issue in medical datasets.
• Experimental results confirmed the model's suitability for automated myocarditis diagnosis.

Conclusions:

• The proposed RLMD-PA model offers a promising automated solution for myocarditis diagnosis using CMR imaging.
• This approach can assist physicians in faster and more accurate patient triage.
• Deep reinforcement learning shows potential for advancing computer-aided diagnosis in cardiology.