胸部X線写真における肺結節状陰影の形状特徴量分析による偽陽性陰影の除去

概要

論文の詳細を見る
We have been developing an automated scheme of computer-aided diagnosis(CAD)for detection of lung nodules based on a difference-image approach in which a nodule-suppressed image is subtracted from a nodule-enhanced image. To further improve the performance of the scheme, we need to reduce the number of false positives and increase the sesitivity for detection of nodules. In this study, we focused on elimination of false positives of pulmonary nodules. First, we investigated an automated extraction method for nodules in a difference image and calculated various measures related to the shape of nodule candidates. Then we analyzed those measures to reduce false-positive detection. Our results indicated that 94 ROIs(13.5%)in the 696 false-positive ROIs were eliminated by using the method described in this study, whereas no true positive ROIs were eliminated. We conclude that our method based on shape analyses would be useful to eliminate some of the difficult false positives in the current CAD scheme.
社団法人日本放射線技術学会の論文
2001-07-20