Research On Lung CT Image Segmentation Technology

Lung cancer is a common malignant tumor. In recent years, due to the influence ofsmoking and various environmental factors, the incidence and mortality of lung cancer arerising rapidly and died of lung cancer in male patients has been ranked first in the countries allover the world, especially the developed industrial countries. In the study of computer-aideddiagnosis of lung disease, lung CT image segmentation is the most important step and it is thekey of accuracy, stability, automation whose results directly affect the subsequent analysis.In the course of evolution, the traditional level set segmentation method has tore-initialize the level set function periodically. The repeated initialization is not onlytime-consuming, and the evolution slows down. This paper analyses a new variationalformulation for geometric active contours that forces the level set function to be close to asigned distance function, and therefore completely eliminates the need of the costlyre-initialization procedure. The experimental results show that in the process of lung CTimage segmentation, this method can obtain a good result.Watershed transform has lots of advantages, such as, light computational burden, highsegmentation accuracy and fast speed. So it has been widely used in medical image. It usuallypresents the phenomenon of “over-segmentation” due to the noise and partial irregularity ofthe image. In the full study of watershed algorithm, this paper analyses an improvedwatershed algorithm, and the experimental results show that this method achieved a very goodresult, especially for curbing the phenomenon of “over-segmentation” in the process of lungCT image segmentation.Finally, the contour detection algorithm based on the level set and watershed whichcombines the advantages of both segmentation methods can accurately converge to theboundary of the target, and avoid the over-segmentation phenomenon. First the algorithm getsthe broad of the object with level set method, and then extracts the accurate boundary withwatershed method. The experimental results the method can obtain a good result in theprocess of lung CT image segmentation.