Temporal subtraction technique for detection of interval changes on digital chest radiographs by using an elastic matching technique

We have been developing a temporal subtraction technique to assist radiologists in the detection of interval changes of pulmonary abnormalities on digital chest radiographs. By subtracting a previous image from an current one, a temporal subtraction image is obtained, in which the lesions can be enhanced because skeletal structures such as ribs and clavicles are eliminated. Although the quality of the subtraction images obtained with our previous technique is relatively good, minor misregistration artifacts were observed in many cases because of the use of polynomial fitting of shift values. In this study, we employed three new techniques to reduce the misregistration artifacts in the temporal subtraction images. The first was a rib edge matching technique which was used to align the rib structures approximately in the two images. The second was an increase in the number of template and search area regions of interest (ROIs) in the two images. The third was the use of an elastic matching technique for the replacement of the polynomial fitting technique employed in the previous subtraction technique. With the new techniques, the quality of subtraction images was improved significantly. In particular, the number of subtraction images with excellent quality was greatly increased from 37% to 57%.