A Novel Method for Mean Geometry Estimation of Chest Anatomy

In this paper we review a point distribution model for representing the mean geometry and shape variations of the chest anatomy from JSRT training set. PDM is based on landmark points which are annotating point placed by human expert. JSRT database includes 247 images of chest radiographs from fourteen institutes in Japan and US. In order to determine the boundaries of the clavicles and heart, we used SCR database which has been labeled by a radiologist and a computer software engineer. 50, 44, 26, 23, and 23 landmarks have been used for left lung, right lung, heart, left clavicle, and right clavicle respectively.The proposed model can be used for initial position of chest organs in automatic medical image segmentation method, which is the first component of any Computer Aided Diagnosis system. The model estimation process is shown to be a quite simple and has much less complexity as compared to the original shape data.

Keywords:

Mean Geometry , Point distribution Model , Chest Organ , Computer Aided Diagnosis , image segmentation