The whole brain segmentation method based on magnetic resonance T1 imaging and its clinical application have been extensively studied. With its advantages in imaging of white matter nerve fiber tracts, the theory and application of diffusion sensor imaging (DTI) has developed rapidly in recent years, and the study of brain segmentation for diffusion tensor imaging has become a problem that must be solved. Image registration is an important technical step to accurately realize brain segmentation. Traditional registration methods do not take into account the influence of image distortion on registration accuracy. The latest symmetric registration strategy is the best choice to solve this problem. Therefore, considering the characteristics of low resolution of diffusion tensor images and the technological advancement of symmetric registration, an automatic method to partition the whole brain of diffusion tensor images was proposed, and statistical SPSS software was used to compare AD patients and NC control group, and the accuracy of the proposed brain segmentation method was verified.