Labeling the pulmonary arterial tree in CT images for automatic quantification of pulmonary embolism

Contrast-enhanced CT Angiography has become an accepted diagnostic tool for detecting Pulmonary Embolism (PE). The CT obstruction index proposed by Qanadli, which is based on the number of obstructed arterial segments, enables the quantification of PE severity. Because the required manual identification of twenty arterial segments is time consuming, we propose a method for automated labeling of the pulmonary arterial tree to identify the arterial segments. Assuming that the peripheral parts of the arterial tree contain most relevant information for labeling, we propose a bottom-up labeling algorithm exploiting the spatial information of the peripheral arteries. A model of reference positions of the arterial segments was trained using manually labeled trees of 9 patients. To improve accuracy, the arterial tree was partitioned into sub-trees enabling an iterative labeling technique that labels each sub-tree separately. The accuracy of the labeling technique was evaluated using manually labeled trees of 10 patients. Initially an accuracy of 74% was obtained, whereas the iterative approach improved accuracy to 85%. The labeling errors had minor effects on the calculated Qanadli index. Therefore, the presented labeling approach is applicable in automated PE quantification.Information and Communication Theory GroupElectrical Engineering, Mathematics and Computer Scienc