Effect of manual image cutout in diagnosis support system of oral mucosal disease based on intraoral image

We have developed a diagnosis support system of mucosal disease which is based on intraoral image. However, the system required the user to cut a image around disease from a intraoral image. It becomes serious problem that cutout images are different from those which are cut out by other users. In this paper, we investigate how different cutout images effect diagnosis results. Four dentists cut out images from 200 intraoral images, and correct classification rate is discussed. From the results, classification rate of the brief career dentist is less than the other dentists, and instruction for image cutout is important for accurate diagnosis.