Abstract: The precise localization of the target structures is the key issue of the treatment of Parkinson's disease.However,these structures were mostly small,blur,and undistinguishable in MR images.In this paper,the segmentation of the target nucleus in Parkinson's disease was studied preliminary.Through constructing the dependence tree model,we segmented the substructures associated with Parkinson's disease.Theory of fuzzy connectedness was used in the segmentation,and then morphology algorithm was used to optimize the results,so that the segmentation result was more desirable.Through compared with the manual segmentation results from hospital experts,the similarity index of algorithm segmentation is over 80%,which can satisfy the clinical requirements.