Lung Cancer Emphasis with Synthesizing Pseudo-normal Image for CAD System

Abstract

A new computer aided diagnosis (CAD) system for a plain chest radiograph is proposed in this study. CAD systems enhancing lung nodules have been actively studied to distinguish the lung nodules. Currently, temporal subtraction (TS) based systems are used extensively for CAD systems to enhance the lung nodules. The TS method can suppress false positives comparatively because it uses the past chest radiograph of the same person. However, the TS method cannot be applied to initial visitors because they have no past chest radiographs. Although a method enhancing the lung nodules with only one chest radiograph selected among other people was proposed, they have several problems. The purpose of this study is to develop a new CAD system which does not need the past chest radiograph. In our method, a pseudo-normal image is synthesized from a database containing chest radiographs of other patients that have already been diagnosed as normal by some medical specialists. Next, the pseudo-normal image is fed into the CAD system instead of a unique past chest radiograph. At last, subtracting the synthesized normal image from target chest radiograph emphasizes lung nodules. The effectiveness and the validity of the proposed method are verified by some experiments.