Enhancement of anomalies for temporally sequential chest radiography. Automated image registration and temporal subtraction technique

Current radiographs were compared with previous ones, and their anomalies were enhanced by the digital subtraction of both images. This procedure affords easy recognition of temporal changes. The image by which the first derivative filter limits the orientation of gradients was used in order for ''lung markings'' to be matched when making adjustments prior to the subtraction of the two images. Previous and current images were then matched using the sequential similarity detection algorithm (SSDA) method. With our method, although artifacts due to disagreement of the costal shadow were generated, subtle opacities such as infiltrates, nodules, bands, linear shadows, and cardiomegaly were clearly enhanced on the subtraction image. This suggests improved accuracy when radiologists diagnose a sequential chest image using the subtraction image. (author)