Semi-automatic detection of scoliotic rib borders using chest radiographs.

ABSTRACT

Stereoradiography is a well known technique to obtain 3D reconstructions of the rib cage. However, clinical applications are limited by the associated 2D rib detection method. Either this detection is widely supervised and time-consuming for the user, or it is fully automatic and not accurate enough for proper 3D reconstruction or clinical indices extraction. To address these issues, we propose a novel, semi-automated technique for detecting scoliotic rib borders in PA-0 degrees and PA-20 degrees chest X-ray images, using a modified edge-following approach. The novelty consists in following multiple promising edges simultaneously. Detections are initiated from starting points (input by the user) along the upper and lower rib edges and the final rib border is obtained by finding the most parallel pair among the detected edges. Promising results show the superiority of this approach over classical rib detection in terms of accuracy. Moreover, the proposed method is of great relevancy in the scoliotic context since scoliotic ribs present very few shape priors, due to their irregularities, and hence, standard rib detection techniques become unsuitable.