Brain Magnetic Resonance Imaging Brain Tissue Segmentation And Multiple Sclerosis Damage Algorithm

Brain MRI (Magnetic Resonance Imaging, MRI) are widely used in medical imaging field because of its high-resolution, high speed, non-ionizing radiation on the human bodies,etc. Among them, the brain MR image segmentation is an important part of automatic computer-aided diagnosis. Accurate segmentation has a crucial role for medical diagnosis and treatment planning. MRI image segmentation is mainly used to extracte interesting target areas in brain MR images, and then to conduct target measurement and analysis. Due to the bias field and the noise and the traditional fuzzy C-means (Fuzzy C-Means, FCM) clustering algorithm using only the pixels of the gray information without spatial information, so segmentation results obtained can not be effectively guarantee the accuracy and robustness. In this paper, we study bias field correction and segmentation of the improved FCM method and multi-modal medical image fusion for the detection of multiple sclerosis (MS).The primary work and remarks of this paper are follows:(1)Use the local strength of clustering with spatial location of kernel-related weighting. We firstly propose a kernel-based spatial information tissue clustering and the bias field correction cost function, and then derive a new iterative optimizing algorithm. Compared results show that our method can overcome the impact of noise data and bias field as well as effectively preserve the structure information.(2) By combining Brain Web brain atlas (atlas) priori information,we establish a FCM cost function model with joint registration and segmentation and proposed a new robust segmentation method. The experiment results show that the improved FCM method based on the guidance of atlas prior information can better overcome the impact of the noise and improve the accuracy of image segmentation. If a brain MR image segmentation contains 7% noise,100% bias field, segmentation accuracy rate will be 90%.(3) We developed an automatic algorithm for MS lesions segmentation by utilizing the fusion T1 and T2-weighted MR brain images based on D-S evidence theory and the improved FCM clustering algorithm. First,segmented T1 and T2-weighted MR brain images by the improved FCM clustering algorithm.Then fused the resultant images according to the joint mass of T1 and T2-weighted MR brain images to produce the segmentation of MS lesions.The experiment Results on MR brain images show that the proposed algorithm is able to improve the segmentation accuracy,which is important to assist the diagnosis of MS in clinic. (4) With the combination of a kernel-based spatial information tissue clustering and the bias field correction cost function and extraction of multiple sclerosis lesions, we made a brain MR image segmentation and multiple sclerosis tissue detection system.