The same subject can be accurately aligned for effective interpretation of two observation images representing a change in the subject and not adjacent to each other in three or more images arranged in temporal order of the change. The subject in one of the observation images is aligned to the subject in a transitional image adjacent to the observation image, and an aligned image is generated. The same procedure is repeated between the previous aligned image and a transitional image adjacent to the previous transitional image to generate another aligned image. A final aligned image is generated by aligning the subject in the observation image to the subject in a reference image. This procedure is carried out for the other observation image and a final aligned image is also generated by alignment to the reference image. A diagnostician compares the two final aligned images.