Research On Fast Segmentation Algorithms Of Medical Image Based On Level Set Method

The technology of image segmentation is significant in medical image processing.Accurate segmentation result could help doctors to estimate the position, size,geometrical shape and the spatial relationship with the surrounding tissues of thelesions. It could be very important in clinical diagnosis, preoperative planning andmedicare design. Chan-Vese model is a classic segmentation algorithm which is basedon the contour, and has been the research highlight of the researchers for quite a longtime. The core of Chan-Vese model is that the energy traces evolutes towards the targetborder during the progress of minimization. So its stopping function only depends onthe information of the grey value. C-V model has adaptive ability for the variation ofthe structure of the contour curve. It makes C-V model detect the target contourautomatically, and the initial contour could be selected flexibly. However, when thevariation of the grey value around the boundary is alleviated, C-V model would comewith the problem of incomplete segmentation because it is excessively depend on theinformation of grey value.This paper made a study on the principle of C-V model and indicated the reasonof the incomplete segmentation. It is the assuasive variation of the grey value aroundthe boundary making the C-V model stop evolution untimely. Based on the studyabove, the paper proposed the C-V model combined with threshold value segmentationmethod. Threshold value segmentation could exclude most of the background,reducing the pixels calculated by the C-V model, which can improve the segmentationspeed. And the range of the grey value would diminish after the segmentation ofthreshold value method. The evolution ability of C-V model would be improved by thechange of scale. So the C-V model could reach the real boundary.Based on the research of the Level Set method, a new level set method combinedwith the fuzzy C-means clustering algorithm was proposed in this paper. FCM methodcould reduce the evolution area of the level set method effectively by the preprocessingof the image. The initial level set got by the FCM method is almost close the edge ofthe target area. The segmentation ability at the narrow edge of the improved methodwould be enhanced by combining with the GVF method.Finally, according the analysis of the characteristics of those two algorithms,preliminarily proposed a selection strategy of segmentation algorithms. The strategy is to judge the characteristics of the images, and then to find an appropriate segmentationalgorithm by matching with the sample. And use Matlab to realize the algorithm above.The comparison results show that the segmentation time of the C-V method based onthreshold value method sharply reduced. It could continue the evolution even at theedge where the variation of the grey value is assuasive to avoid the incomplete; Thelevel set method combined with the FCM method could address the defect of the levelset method, or it would initialize the level set continuously. So the follow-up effect ofthe level set method is more accurate.