Left Ventricle wall extraction in cardiac MRI using region based level sets and vector field convolution

Left Ventricle imaging using short-axis MRI sequences is considered as an important tool used for evaluating cardiac function by calculating important clinical cardiac parameters. This requires manual tracing of LV wall which is subjective, tedious and time-consuming process. This paper presents semi-automatic method for left ventricle inner wall (endocardium) segmentation. This paper focuses on segmenting one complete cardiac cycle without any user intervention. The method used in this paper is region based level sets and vector field convolution active contour model out of which the later method has significantly achieved the better segmentation results. The end systolic and end diastolic volume is calculated by both the methods. The methods are tested on many images and time consumption is reduced using vector field convolution which takes only 30 iterations for segmenting one image per slice. The clinical parameters end diastolic volume, end systolic volume and ejection fraction values obtained from both methods are compared with the values of manually segmented images. The value obtained from vector field convolution gives a closer value to manual segmentation which proves the accuracy of the method and can be considered clinically significant. This semi-automatic approach provides cardiac radiologists a practical method for an accurate segmentation of left ventricle.