An Automatic Pulmonary Nodules Detection Method Using 3D Adaptive Template Matching

This paper proposes a fast and robust computerized scheme for automatic detection of pulmonary nodules by using 3-dimensional adaptive template analysis. First, lung lobes are segmented clearly via the 3D region growing algorithm and then a number of regions of interest (ROIs) are extracted by using OTSU threshold segment algorithm. Second, to establish adaptive template for every ROI, a novel template matching technique is applied. Third, the similarity between the 3D adaptive template and pulmonary nodule candidate image is calculated by utilizing Normal Cross Correlation (NCC) algorithm. Finally, data, which are collected from Lung Image Database Consortium (LIDC) image dataset and Shanghai Lung Hospital, are selected to validate the efficiency of this algorithm. Results show that this proposal algorithm can detect the pulmonary nodule accurately. The sensitivity of former datasets is 95.29% and the false positive rate is 12.90%, while the sensitivity of later datasets is 87.82% and false positive rate is 18.68%. In conclusion, the underlying algorithm is effective and robust.