Fully automatic segmentation of 4D MRI for cardiac functional measurements

Area of Science:

• Medical Imaging
• Cardiovascular Imaging
• Image Segmentation

Background:

• Cardiac function measurement relies on accurate segmentation of cardiac magnetic resonance (MR) images.
• Fully automatic segmentation of the left ventricle (LV), right ventricle (RV), and myocardium in 4D cardiac MR images remains a significant challenge.

Purpose of the Study:

• To propose a novel deformable-based segmentation methodology for efficient and accurate segmentation of 4D cardiac MR images.
• To automatically segment the left ventricle (LV) and right ventricle (RV) in 4D cardiac MR images.

Main Methods:

• The methodology utilizes the Hough transform and local Gaussian distribution (LGD) for LV endocardial contour segmentation.
• A level set-based shape prior method is employed to delineate LV epicardial contours and the RV boundary.

Main Results:

• The automatic segmentation approach was applied to 17 subjects, demonstrating efficiency compared to manual segmentation.
• Achieved high segmentation accuracy with average Dice values: 88.62% (LV endocardial), 87.35% (LV epicardial), and 82.63% (RV contours).

Conclusions:

• A method for accurate LV and RV segmentation in 4D cardiac MR images has been presented.
• The proposed method outperforms existing techniques, yielding superior Dice values for LV segmentation.
• This approach is suitable for clinical evaluation of ventricular volume, size, and thickness.