Automatic Segmentation of Mitral Valve for Detection of Rheumatic Heart Disease using Ultra Sound Images

Abstract

Rheumatic Heart disease is an outcome of continuous episodes of rheumatic fever. It directly damages heart valve especially mitral valve, which leads to thickened leaflet and fused tips. So results in mitral stenosis a condition with narrowed valve opening which results in reduction of flow of blood. Echo-cardiogram provides valuable information about heart condition for exact and timely detection of heart disease but segmentation of ultrasound images has been considered as a perplexing task due to its intensity combination or low contrast. Manual segmentation is a complex process depends mainly upon the intensity of the images but often it includes local information too. It also requires highly trained team of doctors, radiologists and technicians. The proposed method incorporate automatically initialization of active contour with distance regularized level set segmentation algorithm in order to segment the region of interest. Our approach combines the advantages of level sets that uses the prior knowledge like shape, intensity, contours and edges of object to be segmented. The aim of the task is to segment anterior mitral valve (AML). The results of segmentation is demonstrated quantitatively with automatic curvature calculation, spline modelling, and Skeletoniztion and Dice similarity coefficient (DSC) matrices. Results are compared with the ground truth provided by highly trained manual segmented dataset. Data set used in this task is provided by the Real Hospital Portuguese, in Recife, Brazil. The results are verified using WHF criteria. The accuracy came out to be 82%. Proposed methodology also compared with Piers method.