A Mutual-Information-Based Global Matching Method for Chest-Radiography Temporal Subtraction

Abstract

Lung cancer is the most common cancer in the world. Early detection is most important for reducing death due to lung cancer. Chest radiography has been widely and frequently used for the detection and diagnosis of lung cancer. To assess pathological changes in chest radiographs, radiologists often compare the previous chest radiograph and the current one from the same patient at different times. A temporal subtraction image, which is constructed from the previous and current radiographs, is often used to support this comparison work. This paper presents a Mutual-Information (MI)-based global matching method for chest-radiography temporal subtraction. We first make a preliminary transformation on the previous radiograph to make the center line of the lungs in the previous radiograph coincide with that of the current one. Then, we specify areas of the lungs to be used for mutual information registration and extract rib edges in these areas. We transform the rib edge image of the previous radiograph until mutual information between the rib edge image of the previous radiograph and that of the current radiograph becomes maximal. Finally, we use the same transform parameters to transform the previous radiograph, and then use the current radiograph and the transformed previous radiograph to construct the temporal subtraction image. The experimental result demonstrates that our proposed method can enhance pathological changes and reduces misregistration artifacts.