Image feature analysis and computer-aided diagnosis in digital radiography: automated delineation of posterior ribs in chest images

Abstract

In order to facilitate computerized quantitative analysis of digital chest radiographs, an automated method for accurate delineation of posterior ribs in frontal chest images is being developed. This method is based on an analysis of vertical profiles in the lung regions and a statistical analysis of edge gradients and their orientations in small selected regions-of-interest (ROIs). A shift-variant function is fitted to vertical profiles to obtain initial estimates of locations of rib edges. Rib edges are then determined more accurately by analyzing cumulative edge gradients and their orientations in small ROIs that are located adjacent to the initially estimated edges. The present computerized method can achieve a good agreement between the detected and the actual rib structures for posterior ribs in 74% of 50 cases examined. This suggests that automated detection of posterior ribs by a computerized method is feasible, and may be useful for computer-aided diagnostic schemes in the chest.