CineMyoPS: Segmenting Myocardial Pathologies From Cine Cardiac MR.

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Medical Image Analysis

Background:

• Myocardial infarction (MI) is a major global health concern, necessitating accurate risk stratification and prognosis.
• Late gadolinium enhancement (LGE) and T2-weighted cardiac magnetic resonance (CMR) imaging are standard for detecting myocardial scarring and edema, but require contrast agents and time.
• Cine CMR offers a rapid, contrast-free method to visualize acute MI-induced myocardial motion and structural changes.

Purpose of the Study:

• To develop and validate a novel deep neural network, CineMyoPS, for segmenting myocardial pathologies (scars and edema) exclusively from cine CMR images.
• To leverage cine CMR's inherent motion and anatomical information for MI assessment, bypassing the need for contrast-enhanced sequences.
• To improve the efficiency and accessibility of cardiac magnetic resonance imaging for MI evaluation.

Main Methods:

• An end-to-end deep neural network (CineMyoPS) was designed to process cine CMR images.
• The network extracts both motion and anatomical features relevant to MI.
• A consistency loss function was implemented to enforce joint learning of interdependent features, inspired by co-training.
• A time-series aggregation strategy was employed to integrate cardiac cycle information for enhanced segmentation.

Main Results:

• CineMyoPS demonstrated promising performance in segmenting myocardial pathologies (scars and edema) from contrast-free cine CMR images.
• The method achieved accurate motion estimation and myocardial anatomy segmentation.
• Experimental validation was conducted on a multi-center dataset, confirming the model's robustness.

Conclusions:

• CineMyoPS effectively segments myocardial scars and edema using only cine CMR, offering a faster and contrast-free alternative to traditional LGE and T2-weighted imaging.
• The deep learning approach integrates motion and anatomical data across the cardiac cycle for improved diagnostic accuracy in acute MI.
• This technique has the potential to streamline MI assessment, risk stratification, and prognosis evaluation.