A method for automated extraction of aorta and pulmonary artery in the mediastinum using medial line models from 3D chest X-ray CT images without contrast materials

Proposes a method of automated extraction of aorta and pulmonary artery (PA) areas in the mediastinum from uncontrasted 3D chest X-ray CT images. The proposed method does not extract contours of these blood vessels directly, but extracts the medial line of each, vessel and recovers each vessel area. First, the process performs edge detection based on the local standard deviation to get edge areas of vessels. Second, the Euclidean distance transformation is applied for non-edge areas and the likelihood image of the center of vessels is obtained. Medial line models are deformed based upon the likelihood image so as to be fit to the center of each artery. The aorta and the PA areas are obtained by applying the reverse distance transformation to medial lines extracted above. We applied the proposed method to seven cases of uncontrasted 3D chest X-ray CT images. The experimental results showed that the aorta and the PA areas could be extracted satisfactorily.