ACTIVE DETECTOR DUAL ENERGY CHEST RADIOGRAPHY CAD

DESCRIPTION (provided by applicant): The specific aim of our proposed research is to develop an improved, computer-assisted detection and characterization (CAD) system for lung nodules in chest x-ray images. The overall objective is to improve the performance of chest radiography so it will provide a viable lung cancer screening method while retaining its low dose, ease of use and low cost. According to a recent paper, "Today, lung cancer is most often diagnosed on the basis of symptoms of advanced disease or when chest x-rays are taken for a variety of purposes unrelated to lung cancer detection. Unfortunately, in the United States no society or governmental agency recommends screening, even for patients with high risks, such as smokers with airflow obstruction or people with occupational exposures, including asbestos." Our approach provides significant technological innovation in two respects. First, unlike almost all previous work, we will use dual energy images. These provide data, unavailable from conventional radiographs, with significant advantages for CAD. It can be used to compute separate images of body structures with different composition. For example, one image contains only soft-tissue structures while another displays only structures containing high atomic number materials, such as calcium. The soft tissue image eliminates rib shadows, which are a principal source of errors in chest radiography CAD. The high atomic number tissue image can be used to compute a sensitive indicator of whether a nodule contains calcium, an important factor in a benign vs. malignant diagnosis. The conventional radiograph is also available so its data can be combined with the dual energy data to improve overall accuracy. A second innovation is the use of a new type of dual energy detector, which we call an "active detector." The active detector uses a new electro-optical approach to provide data with a much larger difference in average energy than previous dual energy detectors. As a result, clinical quality images can be made with one-fourth the dose of the current state-of-the-art. In Phase I, we will develop and implement a first version of an active detector dual energy CAD system. We will show its feasibility with existing images made with our prototype installed at the University of California, Davis Medical Center.