Abstract
To understand pathology changes (generation and progress) in chest diagnosis, radiologists often compare multiple chest radiographs obtained from the same patient in different periods. To support such a work, temporal subtraction images constructed from the previous and current radiographic images with the enhanced visualization of pathology changes are expected. This paper presents a genetic-algorithm-based method for constructing temporal subtraction images. First, we extract the corresponding rib images from the previous and current images, and use them for global matching of the two images. Then, we divide the lung area in the current image into many subareas, and for each subarea, we use the genetic algorithm for local matching to find an optical corresponding area in the previous image efficiently. The experimental results show that in the constructed temporal subtraction images, pathology changes are enhanced and only few misregistration artifacts arise. Thus, our method is useful for constructing temporal subtraction images.
