Image feature analysis for computer-aided diagnosis: accurate determination of ribcage boundary in chest radiographs

A computerized method for accurate determination of the ribcage boundary in chest images has been developed for use in computer-aided diagnosis (CAD) schemes for automated detection of abnormalities such as the pulmonary lung nodules, pneumothorax, interstitial disease, cardiomegaly, and interval changes in clinical chest images. With our method, the average position of the top of the lung in the chest image is determined first. Top lung edges and ribcage edges are determined within search ROIs, which are selected over top lung cages and ribcages. Three polynomial functions are applied separately to yield smooth curves for top lung edges and right and left ribcage edges. The complete ribcage boundary is then obtained by smoothly connecting three curves. A total of 1000 radiographs were digitized to 1k x 1k matrix size and a 10-bit gray scale with a laser scanner and analyzed by our method. The subjective evaluation indicated that our method produced moderately to highly accurate results in approximately 96% of the 1000 cases examined.