Improving Nodule Detection in Chest CT Images Using a Cylindrical Filter Based on the Anatomical Structure of the Lung

概要

論文の詳細を見る
In a previous study, we developed a cylindrical filter for the detection of lung nodules in chest CT images. However, the detection rate of this method is reduced when nodules are adjacent to or overlap blood vessels or bronchi. The main objective of the present study was to develop a novel technique for improving the nodule detection capabilities of our method based on the anatomical structure of the bronchi and blood vessels obtained from CT images. In the proposed method, chest CT images are divided into a region consisting of normal lung structures (bronchi and blood vessels) and another region. Then, cylindrical filters with different characteristics are applied to the two regions. In order to evaluate the effectiveness of this method, the proposed method and the conventional method were applied to chest CT images, and their nodule detection capabilities were compared. The results showed that the true positive rate of the conventional method was 0.72, whereas that of the proposed method was 0.79. The number of false positives per case was 4.19 for both methods. These results indicate that the proposed method may be useful for improving nodule detection performance.