Research On Lung CT Image Segmentation Based On Graph Cuts And Level Set

Lung CT image segmentation is the technology that extracts lung tissue images from the background image. Because segmented organization will not only help reduce the burden on doctors, but also is the infrastructure of the follow-up and more advanced applications such as visualization and reconstruction. So how accurate and fast to complete segmentation is a hot study point. Meanwhile, because of the characteristics of noise, low contrast, fuzzy boundaries, dense texture and other in lung CT image, it is also difficult to improve the accuracy as well as the efficiency of lung CT image segmentation.According to with human interaction or not, current medical image segmentation methods can be divided into semi-automatic segmentation and automatic segmentation method. Because of combining the operator’s knowledge and the processing power of the computer, semi-automatic segmentation method usually gets ideal segmentation results, thus often used in clinical. The current study on this kind of method focus on improving the interaction friendliness and the segmentation efficiency. On the other hand, The automatic segmentation method, as the future of Lung CT image segmentation technology, can be completely operated without human intervention. But due to the complexity of medical images, automatic segmentation methods often cannot obtain satisfactory results, it is mainly still in the laboratory stage. Its research focuses on improving the accuracy of segmentation and segmentation speed. In this paper, based on the research of the characteristics of lung CT image, combining with the advantages and disadvantages of level set and graph cuts, we propose a semi-automatic segmentation algorithm and an automatic segmentation algorithm, the main innovations and contributions are as follows:① Proposed lung CT image semiautomatic segmentation algorithm based on graph cuts. The algorithm improves the efficiency of graph cuts algorithm, using superpixels algorithm SLICO(zero parameters version of simple linear iterative clustering) to divider the image into many small blocks, which can replace the pixels as node to build graph and sharply reduce the number of nodes. Besides, our method optimizes interaction based on shape features and improves the weight formula of side,while using simplified Grab Cut to segment the image iteratively. With all the improvements, our method has better iteration and fast speed in image segmentation.② Proposed lung CT images automatic segmentation algorithm based on level set. For sake of solving the disadvantages of complexity, more iterations and inaccuracy border of level set, our method combines the breadth-first search and morphological operation, acquires the general outline of the lung area in advance, based on lung CT image features. Then, the method uses the level set technology with a fewer number of iteration to obtain a more precise target contour boundary, and finally, using graph cut algorithm to optimize the boundary contours. As a result, with respect to the normal level set algorithm in the case of automatic segmentation, our method improves the speed and accuracy of segmentation segmentation.