Publications

Abstract : The assessment of cardiovascular disorder acuteness is of great concern worldwide to provide appropriate therapeutic interventions. 3D models aid the examination of complex heart anatomy and orientation that could improve surgical planning. In this work, an attempt is made to develop a computationally efficient framework for reconstruction of 3D left ventricle (LV) models to assess the severity level of cardiovascular abnormality from segmented cardiac magnetic resonance (CMR) images. The novelty of this work relies on the reconstruction of 3D LV models in different dimensions to measure the significant variations in cardiac abnormality. The short-axis view CMR images for healthy, mild, moderate and severe abnormal subjects are obtained from second Annual Data Science Bowl database. Initially, the LV is segmented in all the slices and 3D models are reconstructed. Here, 600 surface models of ventricle have been created from 9600 2D slices of 20 subjects. The measured end-diastole and end-systole volume is correlated with the manual volumes provided in the database. It also presents the assessment of cardiovascular disorder severity based on variations in ventricular volume over a cardiac cycle. It is observed that, the calculated volumes correlates with the manual volumes. The performed study reveals that variation in cardiac volume indicates the level of deformation in ventricular chamber in a cardiac cycle. This study shows the potential usefulness of 3D reconstructed LV models in the understanding of heterogeneous ventricle anatomy and discrimination of different categories of cardiac abnormality. Thus, this proposed frame of work can assess the heart functionality that could assist the radiologist in the diagnosis of severity of cardiovascular disorder.