Automated detection of chest nodules in 3D chest CT scans by using 2nd-order autocorrelation features

概要

論文の詳細を見る
Lung cancer screening using low dose X-ray CT scan has been an acceptable examination to detect cancers at early stage. We have been developing an automated detection scheme for lung nodules on CT scan by using second-order autocorrelation features and the initial performance for small nodules (<7mm) shows a high true-positive rate with less than four false-positive marks per case. In this study, an open database of lung images, LIDC (Lung Image Database Consortium), was employed to evaluate our detection scheme as an consistency test. The detection performance for LIDC, including the first data set opened by the consortium, was 83% (10/12) true-positive fraction with 3.3 false-positive marks per case.
社団法人電子情報通信学会の論文
2007-01-19