ERS transform for the automated detection of bronchial abnormalities on CTof the lungs

The identification of bronchi on Computed Tomography (CT) images of the lun gs provides valuable clinical information in patients with suspected airway s diseases including bronchiectasis, emphysema, or constrictive obliterativ e bronchiolitis. The automated recognition of the airways is, therefore, an important part of a diagnosis aid system for resolving potential ambiguiti es associated with intensity-based feature extractors. On CT images, near-p erpendicular cross sections of bronchi normally appear as elliptical rings and this paper presents a novel technique for their recognition. The propos ed method, the edge-radius-symmetry (ERS) transform, is based on the analys is of the distribution of edges in local polar coordinates. Pixels are rank ed according to local edge (E) strength, radial (R), uniformity and local s ymmetry (S). A discrete implementation of the technique is provided which r educes the computational cost of the ERS transform by using a geometric app roximation of the intensity patterns. The identification of the adjacent pu lmonary vessels with template matching then allows for the automated measur ement of bronchial dilatation and bronchial wall thickening. Computationall y, the method compares favorably with other methods such as the Hough trans form. Noise-sensitivity of the technique was evaluated on a set of syntheti c images and nine patients under investigation for suspected airways diseas e. Agreement for the automated scoring of the presence and severity of bron chial abnormalities was demonstrated to be comparable to that of an experie nced radiologist (kappa statistics k > 0.5).