3D automatic lung segmentation in low-dose CT

The amount of information generated by medical imaging procedures as well as the number of exams performed all over the world is increasing over time. This leads to the need of faster and more efficient ways to deal with the large datasets characteristic of these procedures. Computer-aided diagnostic methods have an important role in this area. This paper presents a fully automatic method for the identification of the lungs in CT images. The lung regions are identified by a threshold operation as a first step. To separate merged lungs, we apply a sequence of morphological operations. Additionally the trachea and large airways are identified and removed in each slice. The proposed approach was tested in several whole-body CT studies presenting positive results.