Precise Identification and Quantitative Analysis of Right Ventricle Using Hierarchical Intensity Clustering with Markov Random Field for Cardiac Mr Images

Abstract

Precise localization and quantitative analysis of the right ventricle from cardiac magnetic resonance imaging (CMRI) images is imperative for assessing cardiopulmonary and cardiovascular mal-functionalities. Due to its poorly defined borders, precise contouring of the RV in CMRI images continues to be difficult. An approach for contouring the right ventricle based on a hierarchical intensity-based clustering method followed by Markov random field is proposed in this paper to overcome this difficulty.Because our method offers localization for each time step, it enables comprehensive right ventricle analysis during the whole cardiac-cycle. It also allows automatic prediction systems of volumetric parameters such as end systole and end diastolestages.48 human participants' cardiac MRI scans were used to validate the method. The presented scheme yielded significantly less variance than (approximately one half) compared to the reference standard of manually determined RV contours by clinical experts. This approach obtained mean Dice-Coefficient and Haussdorff-Distance of 0.92 and 5.25 mm 0.94 and 5.68 mm on the validation and tests. Further the results are evaluated using diagnosis metrics such as end diastole volume, end systole volume, and ejection fraction. Our model heading towards accurate RV localization at endocardium borders in cardiac MRI.