MyoNet: Deep Learning-Based Myocardial Strain Quantification from Cine Cardiac MRI

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Biomedical Engineering

Background:

• Assessing myocardial regional function from cine cardiac magnetic resonance (CMR) images is crucial for diagnosing cardiac conditions.
• Existing methods for myocardial strain analysis require efficient and accurate alternatives for pre-clinical and clinical applications.

Purpose of the Study:

• To develop and evaluate MyoNet, a deep learning (DL) network for measuring myocardial regional function from cine CMR images.
• To compare MyoNet's performance against ResMyoNet and SinMod-derived reference strains.

Main Methods:

• MyoNet and ResMyoNet were developed using DL, employing advanced convolution operations for spatial and temporal analysis of cine CMR images.
• Both networks were optimized for detailed myocardial deformation and utilized robust loss functions.
• Performance was assessed on datasets from Dahl salt-sensitive rat models undergoing radiation therapy (RT).

Main Results:

• MyoNet demonstrated superior performance in myocardial strain measurement, showing high consistency with SinMod-derived reference strains.
• MyoNet achieved higher performance metrics than ResMyoNet, including SSIM (0.961/0.960), ICC (0.973/0.975), and Pearson CC (0.973/0.953) for circumferential (Ecc) and radial (Err) strains.
• Statistical analyses validated MyoNet's accuracy and efficiency in generating strain measurements.

Conclusions:

• MyoNet represents a significant advancement in myocardial strain analysis from cine CMR images.
• Its accuracy, efficiency, and reliability position it as a valuable tool for pre-clinical studies and clinical applications, especially for monitoring cardiac health in cancer patients.