A New Pulmonary Nodules Computer-Aided Detection System in Chest CT Images Based on Adaptive Fuzzy C-Means Technology

This paper presents a new pulmonary nodules computer-aided detection system in chest CT images utilizing the adaptive fuzzy C-Means (AFCM) technologies. Since rough segmentation of nodules tends to result in high false positive (FP), the main purpose of this study is to reduce the false-positive of candidate nodules via the clustering and classifying approaches. The proposed scheme consists of three phases: pulmonary nodule identification, training nodules clustering, and testing nodules classification. Firstly, the lung parenchyma is extracted through neighborhood connected technology and masking processing, and by appropriate thresholding processing, the candidate nodules are identified. Then, for improving the performance in the training phase, we utilize the AFCM technology. Finally, the category of each testing candidate nodule is determined by Mahalanobis distance. We validated our method on 35 volumes of chest CT, which is subdivided into 20 training part and 15 testing part, and an approximate false-positive of 2.8 per scan is obtained in our experiment. The preliminary results prove that our scheme is a promising tool for pulmonary nodule detection.