Automatic Segmentation of the Aorta in Cardiac Medical Images

Abstract

Automatic aorta segmentation from three dimensional image data is very important for treatment planning and diagnosis of cardiovascular diseases, which are the leading cause of death in the developed countries. A number of techniques are available but most of them need user interaction and require very high contrast of input images that increase the patient exposure time inside the machine. In this paper, we have proposed a novel segmentation pipeline that can automatically segment the aorta from low contrast three-dimensional data. Input images are pre-processed using gradient and sigmoid techniques in order to identify real edges and increase contrast of the image. Hough transform is used for automatic detection of circular shape of aorta that is followed by a three dimensional connected threshold algorithm to delineate the aorta from other objects. The results show that our technique can automatically segment the aorta faster than existing techniques. The algorithm uses low contrast image data, requires no user interaction and no parameters tuning. The proposed technique, with minor modifications, can also be used for semi-automatic segmentation of other organs in the human body.