In this study, a technique for computer-aided diagnosis (CAD) systems to detect lung nodules in X-ray pulmonary computed tomography (CT) images is proposed. The adaptive border marching algorithm was implemented for lung volume segmentation. Region growing and rule based method were used to detect the nodules candidates. Then, we extracted a total of 11 features, including intensity features and geometry features, of these candidates. The fuzzy min-max neural network classifier with compensatory neurons (FMCN) was advanced by K-means clustering, for false-positive reduction. In hyper-space, the cluster is similar to hyperbox, thus the K-means clustering algorithm was implemented for determine the expansion coefficient (hyperbox size). Nineteen clinical cases involving a total of 5766 slice images were used in this study. 26 nodules out of 31 were detected by our CAD (the sensitivity about 84%), with the number of false-positive at approximately 2.6 per CT scan. The preliminary results show that our scheme can be regarded as a potential technique for CAD systems to detect nodules in pulmonary CT images. Keywords-Computer-aided diagnosis; lung nodules; K-means cluster; fuzzy min-max neural networkI.