Viewpoint Recognition in Cardiac CT Images

Position and orientation information is often lacking in DICOM datasets. This creates a need for human involvement or computationally expensive 3D processing for any analytical tool, such as a software-based cognitive assistant, to determine the viewpoint of an input 2D image. We report a solution for cardiac CT viewpoint recognition to identify the desired images for a specific view and subsequent processing and anatomy recognition. We propose a new set of features to describe the global binary pattern of cardiac CT images characterized by the highly attenuating components of the anatomy in the image. We also use five classic image texture and edge feature sets and devise a classification approach based on SVM classification, class likelihood estimation, and majority voting, to classify 2D cardiac CT images into one of six viewpoint categories that include axial, sagittal, coronal, two chamber, four chamber, and short axis views. We show that our approach results in an accuracy of 99.4 % in correct labeling of the viewpoints.