Automatic segmentation of right ventricular ultrasound images using sparse matrix transform and a level set

Area of Science:

• Medical Imaging
• Computational Cardiology
• Biomedical Engineering

Background:

• Accurate segmentation of the right ventricle (RV) in echocardiography is crucial for assessing cardiac function.
• Manual segmentation is time-consuming and subject to inter-observer variability.
• Existing automated methods may lack precision in capturing complex myocardial motion.

Purpose of the Study:

• To develop and validate an automatic segmentation framework for the right ventricle (RV) in echocardiographic images.
• To accurately segment both epicardial and endocardial boundaries of the RV.
• To provide a reliable tool for quantitative cardiac imaging.

Main Methods:

• A novel framework combining sparse matrix transform, a training model, and a localized region-based level set.
• Sparse matrix transform extracts myocardial motion as eigen-images.
• An RV training model provides initialized segmentation for the level set algorithm.

Main Results:

• The framework achieved high accuracy in segmenting both epicardial and endocardial boundaries.
• Mean Dice coefficients were 90.8 ± 1.7% (epicardial) and 87.3 ± 1.9% (endocardial).
• Low mean absolute distances (2.0 ± 0.42 mm / 1.79 ± 0.45 mm) and Hausdorff distances (6.86 ± 1.71 mm / 7.02 ± 1.17 mm) were reported.

Conclusions:

• The proposed automatic segmentation method is effective for RV analysis in echocardiography.
• The combination of sparse matrix transform and level set segmentation offers robust performance.
• This automated approach can enhance the efficiency and accuracy of quantitative cardiac imaging.