The applicants proposed to develop computer-aided diagnostic (CAD) schemes for detection and characterization of pulmonary nodules in digital chest images. For development of reliable and predictable CAD schemes, they proposed to establish a large database with 2,000 cases of chest radiographs, which include 1,000 nodule cases and 1,000 non-nodule cases, in collaboration with Richard M. Slone, M.D., Mallinckrodt Institute of Radiology, Washington University, under a consortium arrangement. An advanced CAD scheme for detection of lung nodules will be developed by incorporating three subtraction techniques--temporal subtraction, contralateral subtraction and energy subtraction--in order to achieve, on average, a high sensitivity of 80-90% with a low false positive rate of approximately 0.5 per chest image. They would investigate the usefulness of the temporal subtraction technique in increasing the detection of subtle nodules overlapped with ribs and also decreasing the number of false positives due to rib-crossings, when a previous chest image of the same patient is available. Contralateral subtraction, which is a novel technique for removal of peripheral ribs in a single PA chest image, will be examined for enhancement in the detection of overlapped nodules and reduction in the number of false positives. They would also investigate the usefulness of energy subtraction soft-tissue image for improved computerized detection of lung nodules in combination with conventional chest images. In addition to the detection task, they would to develop CAD schemes for characterization of nodules in order to distinguish between benign and malignant nodules. This characterization task is to provide the likelihood of malignance of lung nodules based on quantitative analysis of image features of nodules detected by computer and/or by radiologists. With the high level of detection performance that they expect to achieve, they propose to develop a prototype CAD workstation and carry out a pilot study to examine the clinical usefulness of CAD schemes on detection and characterization of pulmonary nodules.
Showing the most recent 10 out of 75 publications
