Automatic Cardiac MRI Segmentation and Permutation-Invariant Pathology Classification Using Deep Neural Networks and Point Clouds

Abstract Segmentation of cardiac MRI images plays a key role in clinical diagnosis. In the traditional diagnostic process, clinical experts manually segment left ventricle (LV), right ventricle (RV) and myocardium (Myo) to get the guideline for cardiopathy diagnosis. However, manual segmentation is time-consuming and labor-intensive. In this paper, we propose automatic cardiac MRI segmentation and cardiopathy classification based on deep neural networks and point clouds. The cardiac MRI segmentation consists of two steps: ( i ) We use a detector based on you only look once (YOLO) to obtain region of interest (ROI) from the sequential diastolic and systolic MRI. ( ii ) We obtain the segmentation masks from the ROI automatically by a fully convolutional neural network (FCN). Subsequently, we reconstruct 3D surfaces by a simple linear interpolation method, then randomly sample uniform 3D point clouds from the 3D surfaces. From the cardiac point clouds, we perform cardiopathy classification using a cardiopathy diagnosis network (CDN). Experimental results show that the proposed method successfully segments LV, RV, and Myo from cardiac MRI images and achieves comparable results against several existing ones. Moreover, the CDN successfully classifies heart diseases based on point clouds and achieves 92% accuracy on the testing dataset.