Computational analysis of digital chest radiography

Studies have shown that a large percentage of malignant lung nodules detected on chest radiographs are overlooked on previous examination(s). One of the causes for this higher error rate is the superimposition of anatomic structures such as ribs and soft tissues, which obstruct the true visual examination of the underlying soft tissue in the radiographs. Therefore, the dual-energy subtraction method is developed to numerically remove the bone from digital chest radiograph, thus improving diagnostic capabilities. The dual-energy subtraction technique is based on the fact that the attenuation of x-rays is different for different materials (bone versus soft tissues) and different x-ray energies. Hence with images from two different energy levels, one should be able to compute the thickness of the bone and eliminate them from the original image(s) numerically. A number of studies have demonstrated the feasibility of this technique; however, all of these works are based on the global subtraction method. Hence, the signal-to-noise ratio (SNR) in the inter-rib regions of the subtracted images is reduced, which leads to a reduction in the diagnostic capability of the chest radiographs in these regions.; The purpose of this research is to improve the current dual-energy subtraction method by enhancing (or at least maintaining) the diagnostic capability of the whole digital chest radiograph. Our method is a pseudo global/localization approach, achieved by subtracting only the chest regions that are obstructed by the bone structures. The procedures involved can be summarized into the following steps: (1) Decompose the dual energy images into soft-tissue and bone-only images by employing the traditional global subtraction technique. (2) Identify the rib structures by locating the upper and lower rib edges utilizing the bone-only image computed in Step #1. (3) Numerically remove the rib structures identified in Step #2 from the original radiograph(s) by utilizing the rib thickness information calculated in Step #1.; Step #1 is required because the soft tissues in the images are hindering the detection procedure. Also, a new bone equivalent material (consisting of polyvinyl chloride and aluminum) and a soft tissue mimicking material (consisting of water and Lucite) have also been examined in an attempt to offer better approximations for the dual-energy subtraction techniques. Based on the experimental results, the current approach is a good starting point for the identification of bone structure in a digital chest radiograph. It is a good starting point because it uses a unique combination of image enhancement tools and edge detection algorithms that together minimize the problems presented by low SNR images. Further improvement of either contrast enhancement and/or edge detection processes will most likely not improve significantly upon the process used here. Improvement will come from either higher detective quantum efficiency image receptors and/or x-ray sources (both in terms of dual energy separation and photon flux). Consequently, additional work will be required before the technique can be confidently incorporated into a clinical dual-energy subtraction algorithm.