Finding groups of airway tree structures in CT images

In clinical everyday life, thoracic surgeons often observe variants of the lung anatomy that have not been covered in specialist books to date. These can be different bronchial structures - when and where do the airways split off to the individual lobes? - as well as the way in which an artery surrounds the bronchus. These variants are rare, but for a surgeon they are crucial in the planning of an operation. Together with the Department of Thoracic Surgery at Rostock University Hospital and Fujifilm, we have set ourselves the goal of finding and classifying these deviations and giving them to the operating physician.

Our work is based on CT thoracic scans, from which masks have already been extracted - lung lobes, bronchus, arteries and veins. We process these masks into tree structures, try to find patterns and group them. We want to develop a classifier that can automatically assign new patient data to a group. We would like to make this classifier available to surgeons so that they can gain a quick overview of the anatomical characteristics of a patient. Furthermore, we hope to gain an overview of the prevalence of the variations found.