Application of computerized nodule detection in screening programs for lung cancer

To assist radiologists in the detection of small lung nodules the authors are developing a computer vision system. In this paper, the authors applied their new method, which employs linear and nonlinear filters, to mass screening radiographs from a lung cancer detection program. Possible nodules are identified initially be difference-image techniques, with use of both linear and nonlinear filtering methods, and false-positive results are eliminated by feature-extraction techniques. Two hundred 100-mm photofluorographic chest radiographs, including 100 with nodules and 100 normal, which had been obtained in a cancer screening program in Japan, were digitized into a 512 x 512 matrix with a laser scanner. The nodule detection scheme was applied to these digitized images