Recognizing the human anatomical structure is an important process for the CAD system using the 3-D multi-slice CT images. The identified bone frames provide the useful information of human anatomical structure. In this research, we propose a method to identify the structure of chest bone frame by classifying the chest bone frame into 5 parts (spine, sternum, ribs, shoulder blades, and clavicles). In this method, we generate the probabilistic atlases of chest bone structure from pre-recognized structures of bone frames, and make a classification of unidentified bone frames based on the atlases. We applied this method to 53 cases of the 3-D multi-slice CT images, and confirmed its effectiveness from the results.
View full abstract