Abstract: In order to obtain better segmentation results under acceptable computational complexity， the advantages and disadvantages of the existing methods are summarized， and then the lung segmentation method which modifies and integrates the simple methods is proposed. Firstly， the clustering method is used to classify the pixels in CT images into two categories， and the small regions with different gray values are obtained. And then the initial region of the lung is identified. Finally， according to the specific characteristics of the CT image， the boundary is corrected by using the gradient and the gray information of the pixels near the initial lung boundary. The processing results of CT images using the above synthesis method show that the segmentation results can be better when the computational complexity is not too large.

Key words: CT images, pulmonary nodules, integrate, clustering, gradient