A method for automated nomenclature of bronchial branches extracted from CT images

This paper describes a method for automated nomenclature of bronchial branches extracted from CT images. When a medical doctor diagnoses chest CT images, lesions' locations are identified by dominant bronchial branches' names. Also, in the case of real or virtual bronchoscopy, a physician performs bronchoscopy with understanding current locations by branches' names. In the previous method, several branching-pattern models were constructed for automated labeling of bronchial branches extracted from CT images. These models were prepared for each part of the bronchus. An appropriate model was selected by evaluating the average differences of running directions of bronchial branches between the models and the extracted tree structure part by part. However, the previous method selected the wrong model and assigned incorrect names. To solve such miss-labeling, the proposed method removes some models from the candidate models for labeling by checking branching pattern in each branching point during labeling process. Also, the proposed method removes models by checking directions of segmental bronchi and selects a model from remaining models in the labeling process of the right upper lobe. In the experiments using 26 cases of chest CT images, the proposed method was able to correctly assign 90% of segmental bronchial branches.