Ren, Tianbao, Wang, Huanhuan, Feng, Huilin, Xu, Chensheng, Liu, Guoshun, and Ding, Pan
Brain magnetic resonance image segmentation has become a hotspot and a difficult point in the field of medical image segmentation, and its segmentation effect directly affects the later pathological analysis and clinical treatment. For the problem of brain image segmentation, firstly, the image is subjected to pre-processing such as histogram equalization to eliminate irrelevant information in the image and enhance the detectability of the information. Then, through the research and analysis of Fuzzy C-Means(FCM) algorithm, Kernel-based FCM(KFCOM) and Weighted fuzzy kernel clustering(WKFCOM) algorithms are proposed. The WKFSOM algorithm combines the advantages of the two algorithms. It not only uses image space information as prior knowledge, but also can deal with image ambiguity. Finally, the KFCOM and WKFCOM algorithms are used to analyze the MRI images of the brain, and the segmentation effects of various algorithms are quantitatively evaluated by MCR. The KFCOM algorithm has a misclassification rate of 9.03% and the WKFCOM algorithm has a misclassification rate of 6.67%. It can be concluded that the WKFCOM algorithm can accurately segment brain tissue efficiently and unsupervised, and has a good inhibitory effect on noise. This will make it easier to obtain clinical information about the disease and bring great convenience to the clinician's diagnosis. • In this paper, the histogram equalization preprocessing process is applied to medical images to enhance the detectability of relevant information. • The KFCOM and WKFCOM algorithms are compared in this paper. It is concluded that the WKFSOM algorithm can better handle the ambiguity of the image. • The KFCOM and WKFCOM algorithms are used to analyze the brain MRI images, and the KFCOM algorithm is quantitatively evaluated to have a lower error rate. [ABSTRACT FROM AUTHOR]