Improved Fuzzy Connectedness Segmentation Method Of Medical Images

Image segmentation is a key step in medical image processing,since it affects the quality of the medical image in the follow-up steps.However,in the practice of processing MRI images,we find out that the segmentation process involves much difficulty due to the poorly defined boundaries of medical images.In this study,we proposed two improved image segmentation methods based on the theoretical framework of fuzzy connectedness.The first method has involved region growing algorithm combined with multiple seeds selection.In theoretical inference and experiments,our algorithm can effectively overcome many problems when manual selection is used,such as the un-precise result of each target region segmented of the medical image and the difficulty of completion the segmentation when the areas are not connected.From practice,our algorithm can not only improve the completion of the result,but also can improve the integrity of each part of the target area,which has the effect of killing two birds with one stone.The second method use NAM(non-symmetry and anti-packing model)as a preprocessing step for the fuzzy connectedness method.This segmentation algorithm can divide the image into sub-pattern sets with the same gray level,thereby greatly reducing the number of pixels required to participate in the calculation when processing,and reducing the storage space required by the way.In this paper,we describes the feasibility and superiority of the combination of NAM algorithm and fuzzy connectedness method from the theoretical aspect,and proves them through experiments.From the results,we can find that the fuzzy connectedness algorithm with NAM has the advantages of low storage space requirements,fast operation speed,and basically no loss of segmentation effect compared with the traditional fuzzy connectedness method.All in all,it is a good supplement of the theoretical branch of fuzzy connectedness segmentation method of medical images.For testing our proposed methods,some original real images,taken from a large hospital,were analyzed.In order to measurement the segmentation effect,the results have been evaluated with some rules,such as Dice's coefficient,over segmentation rate,and under segmentation rate.The results show that the proposed methods have ideal segmentation boundary on medical images,meanwhile,it has a low time cost.This thesis has a rigorous theory of work and a persuasive test process,in conclusion,it has certain research results.