マンモグラムCADシステムにおける乳房領域自動抽出アルゴリズムの改良

抄録

We have been developing a computer-aided diagnosis(CAD) system for detecting clustered microcalcifications and masses on digital mammograms. In our recent study, we found that the automated breast-region- extraction algorithm used in this CAD system was not accurate sometimes when we increased the database and tested many mammograms. This algorithm was based on the breast-border determination technique. The purpose of this study is to modify and improve this technique and also to test it by increasing image data. In our revised technique, the weight of the location information of breast border and the differentiation of the image density profile were included, and the location of breast region on the film and the number of breasts imaged on the same film were also considered. The new technique was succeeded in the extraction of the breast region in 728 cases(98.8%) out of 737mammograms obtained from 8 different facilities.