Automated detection of lung nodules in chest radiographs using a false-positive reduction scheme based on template matching

Automated detection of lung nodules in chest radiographs is important to reduce false negatives in the diagnoses of lung cancers using chest radiography. The automated nodule detection techniques consist of two steps of initial nodule candidate detection and false positive reduction. In this paper we propose an improved scheme for each of these steps. The proposed false-positive reduction scheme uses template matching technique. As the result of experiments using 125 images with nodules in the JSRT database which is a public database, the proposed nodule-candidate detection scheme gave sensitivity of 96% with 136.5 false positives per image. For evaluation of the total performance of the proposed nodule detection scheme, we created 40 date sets by 40 randomized selection of 80 training images and 45 test images from the 125 images with nodules in the JSRT database. As the result of experiments using these 40 data sets, the proposed nodule detection scheme gave 9.5, 12.5, and 13.8 false positives per image for sensitivity values of 60.2, 69.8, and 74.5% on the average of 40 data sets. The time needed by the proposed nodule detection scheme, excluding the time needed by lung segmentation, was 5.1 seconds per image on the average of 40 data sets using a 3.3GHz Intel PC.