Comparative Approach For Localization Of Right Ventricle From Cardiac Mr Images.

Cardiac MRI modality with excellent contrast plays an important role for both anatomical and functional evaluation of the heart. The localization of the right ventricle is quite a critical process due to the complex shape of the organ and its fuzzy borders than surrounding tissues. In this paper we present our work with two approaches, first one is clustering based and second is deep learning based. The experimentation is performed on short axis cardiac MR images available as heart database by A2SI LABs. The performance of algorithms is evaluated using two statistical parameters Dice metric and Haussdorff distance parameters and ejection fraction. is the quantitative parameter. The Dice metric and Haussdorff distance achieved using both approaches is 87%, 96% and 3.50mm, 2.65mm respectively. The result shows that deep learning approach gives promising results over clustering based method. [ABSTRACT FROM AUTHOR]