Abstract: A novel brain structure segmentation algorithm based on diffusion tensor imaging (DTI), which can extract white matter (WM) is presented. First, DTI diffusion and anisotropy parameters are computed to achieve brain images of each DTI parameter channel. Then, using the expectation maximization (EM) model, WM/non-WM regions in the anisotropy parameters channel are calculated. Finally, using simultaneous truth and performance level estimation (STAPLE) model, segmentation results of each DTI parameter image are fused to achieve the final brain white matter structure segmentation results. Experiments' results showed that the algorithm can accurately segment, WM in brain structures.