Bone suppression in chest radiographs by means of anatomically specific multiple massive-training ANNs

Our purpose was to separate bony structures such as ribs and clavicles from soft tissue in chest radiographs (CXRs). Although massive-training artificial neural networks (MTANNs) have been developed for suppression of ribs, they did not suppress rib edges, ribs close to the lung wall, and the clavicles well. To address this issue, we developed anatomically specific multiple MTANNs that are designed to suppress bones in different anatomic segments in the lungs. Each of 8 anatomically specific MTANNs was trained with the corresponding anatomic segment in the teaching bone images. The output segmental images from the 8 MTANNs were merged to produce a whole bone image. Total variation minimization smoothing was applied to the bone image for reduction of noise while edges were preserved;, then this bone image was subtracted from the original CXR to produce a soft-tissue image where bones are suppressed. We compared our new method with the conventional MTANNs by using a database of 110 CXRs with pulmonary nodules. Our anatomically specific MTANNs suppressed rib edges, ribs close to the lung wall, and the clavicles in CXRs substantially better than did the conventional MTANNs.