Abstract: Brain fiber clustering technologies can effectively segment different levels of fibers bundles and have great vitality during the analysis process.However,the existing fiber clustering technologies usually depend on the spatial location of the fiber tracts and lack the consideration of the brain fiber trend information. This paper proposes a novel brain fiber clustering method based on B-spline fitting and regression model, which performs B-spline fitting, sampling, and standardization of brain fibers, uses linear regression equation to describe brain fibers, con-structs Gaussian density function for each fiber tract and maximum likelihood function for fiber distribution, and then uses EM clustering algorithm for clustering. The algorithm and QB clustering algorithm are applied to real medical data (Brain fibers tracked by PPMI image data). The clustering results are evaluated to qualitatively judge the spatial similarity of the results in anatomical space and quantitatively compare brain fiber tract numbers. The experimental results show that the proposed method can segment brain fiber bundles with anatomical identification at the brain functional region level.