Automated Liver Segmentation Method For CT Dataset Based On Probabilistic Atlas And Shape Model

Medical image segmentation is to separate an image into multiple regions with similarcharacteristics, and propose new technology and processing based on the area that we areinterested. Those are the important method and means we do researches in medical imageanalyzing. In medical image processing and analyzing, image segmentation is certainly animportant part of image analyzing, which is also important in our work on image analyzing. Thefocus of medical image segmentation is to segment the area we are interested in, such as thelesion area and tumor,etc., and make our segmentation result closing enough to anatomicalstructure, so as to help the clinicians analyzing patients’ condition, and making diagnosis andtreatment correctly. Although there are many different segmentation methods, we have not yetdeveloped an effective segmentation method which can satisfy the clinical demands. Due to thedifference of individual liver and structures, combined with the lesion area worse a lot more, theclinical algorithm demands higher requirements on medical image segmentation accuracy. Onthe other hand, the image noise and artifacts affect the image analyzing, making the existingimage segmentation algorithms far away from the clinical demands.Because of these influencing factors mentioned above, the liver medical imagesegmentation has been at the forefront in the field of medical image processing and analyzing. Ithas turned into a research hotspot, and also a challenge. Liver image segmentation of thedifficulty lies in the shape of variety and complexity of its structural texture. At the same time,the blur edges between these surrounding tissues and organs making the accurate segmentationof liver image more challenging.There are many methods applied to image segmentation nowadays, such as region growing,probability map matching, sparse shape composition, level set, etc. as well as many combinedmethods. While the method based on statistics and shape model works best.To deal with the difficult problem of liver segmentation, we propose a combined construction based on probabilistic atlas and graph cut, the information extracted from the atlasprobabilistic and shape prior makes a more comprehensive expression of liver shape.Firstly, we dug into the traditional image segmentation methods, and then combined thevariability of liver shapes, proposed a combination of probabilistic atlas and graph cut, andconducted in-depth researches. This method combined traditional methods and new methods,introduced in the characters based on the liver image and constructed an objective function toachieve a more accurate and fast segmentation on medical liver images.There are two innovations in this paper. The first one is applying SIFT algorithm in featurepoints matching in constructing the probabilistic atlas, which ensures the accuracy of registrationand reduces the amount of calculation. At the same time it improves the efficiency of this newproposed method. The second innovation is applying TV-L1model to reduce the impacts ofimage noises and artifacts before using mean shift algorithm in the process of graph cut.Experimental results show that automated liver segmentation method in this paper for CTdataset based on shape model and probabilistic atlas construction can be used in liver boundarydelimitation, and also to save the liver edge information details. This experimental results showthat this method can meet the efficiency and stability on the basis of improving accuracy.