2D Statistical Shape Model for Lung Using Apex Anatomical Landmark-based Registration Criteria

Statistical Shape Models have been widely used in range of medical applications such as segmentation, registration, classification and interpretation of medical images. Advanced imaging systems such as 4DCT, 3DCT and PET scan have been used to analysis of diseases, however in addition to their high costs, expose the patients to the relatively high radiation doses. Shape representation using routine radiography imaging technique is desire as it does fast and inexpensive along with lower radiation doses. In this work, we propose a novel method to model 2D lung from a small number of Chest X-Ray images. The proposed method combines a contour-based shape descriptors technique and Apex Anatomical Landmark-based (AAL-based) registration criteria to build an optimal 2D statistical shape model of lung which can be used to better analysis of human lungs. The presented model in current work is compared to our prior model. The compactness of the proposed model is evaluated and compared to the prior model. The results show that the proposed model is able to explain more than 90% of total variations using only three principal component modes for both of the right and the left lungs.