マルチデテクタCT画像におけるコンピュータを用いたTracheal Bronchusの自動検出アルゴリズムの開発

概要

論文の詳細を見る
Tracheal bronchus(TB) is a rare bronchial anomaly, which should be non-invasively diagnosed in chest computed tomography(CT) images. However, it can remain undetected because of its low frequency. A small tool to detect this anomaly in chest CT images will be useful to reduce the incidence of cases in which it goes undetected. We developed a scheme for detection of tracheal bronchus consisting of a two-dimensional rule-based scheme that indicates a slice including TB. Criteria applied to identify tracheal bronchus are : in the slice 1) the trachea region is split into two or more regions, and the area of the larger of the largest two regions is four or more times that of the smaller region, and 2) the circularity of the trachea region is lower than 0.65 and recovers in the next slice. Once one of the criteria is satisfied, the slice is determined to include TB. The scheme was applied to 30 cases(three with and 27 without TB) retrospectively. The sensitivity and specificity was 66.7% and 92.6%, respectively.
医用画象情報学会の論文