Lung nodule segmentation using active contour modeling

In this paper, we propose an automatic lung nodule segmentation algorithm using computed tomography (CT) images. The main contribution is automatically detecting large or small non-isolated nodules connected to the chest wall and accurately segmenting solid and cavity nodules by active contour modeling. This method consists of several steps. First, the lung is segmented by active contour modeling. The initialization is the main core of this step. It causes to transfer non-isolated nodules into isolated ones. Then, regions of interest are detected using 2D stochastic features. After that, an anatomical 3D feature is used to detect nodules. Finally, contours of detected nodules are extracted by active contour modeling. At the end, the performance of our proposed method is reported by experimental results using clinical CT images. All nodules (including solid and cavity) are detected and the number of FP is 3/scan.