Endocardial border detection in cardiac magnetic resonance images using level set method

Area of Science:

• Medical Imaging
• Cardiovascular Imaging
• Image Segmentation

Background:

• Accurate segmentation of the left ventricle (LV) in cardiac magnetic resonance (MR) images is crucial for assessing cardiac function.
• Current methods often rely on manual contour tracing, which is time-consuming and subject to inter-observer variability.
• Global measurements like volumes and ejection fraction depend on precise LV segmentation.

Purpose of the Study:

• To develop and validate a novel, automatic method for left ventricle endocardial border detection in cardiac MR images.
• To improve the efficiency and consistency of cardiac function evaluation through automated segmentation.
• To establish a robust segmentation technique that minimizes reliance on manual input.

Main Methods:

• A level set segmentation-based approach was employed for automatic endocardial border detection.
• Image thresholding was used to generate an initial binary mask for the level set algorithm.
• An automatic method evaluating object roundness was utilized for precise localization of the left ventricular cavity to initialize the mask.
• The segmentation process involved initializing and then executing the level set algorithm.

Main Results:

• The automated segmentation method demonstrated high performance in detecting the left ventricle endocardial border.
• Validation against manual contours traced by two experts showed excellent agreement.
• The overall average similarity area between automated and manual segmentations reached 97.89%.

Conclusions:

• The proposed automatic level set segmentation method is effective for left ventricle contour detection in cardiac MR images.
• This automated approach offers a reliable and accurate alternative to manual segmentation, enhancing diagnostic capabilities.
• The high similarity score validates the method's potential for routine clinical application in cardiac function assessment.