Abstract: Segmentation of cerebral infarction focus is an important pre-processing step to assess functional injury of brain． To solve the low accuracy segmentation problem caused by blurred boundaries，irregular shapes and in-homogeneous intensities of cerebral infarctionin the image of diffusion weighted imaging (DWI)，a segmentation method of active contour model based on fuzzy speed function was proposed． Firstly，the Bayesian probability in the wavelet transform domain was used to obtain the initial contour，according to which the real boundary of cere-bral infarction focus can be quickly located，which may improve the segmentation accuracy of the model． Second-ly，the local entropy of the image was introduced into the active contour model． The local entropy of the image can represent water molecule distribution differences in DWI image，thud solve the problems of inhomogeneous intensities and noise interference to a certain extent． Thirdly，according to the fuzzy characteristics of the cerebral infarction fo-cus，a fuzzy clustering algorithm was proposed to calculate the fuzzy membership degree in the fuzzy speed function by combining the local entropy characteristics and gray-scale characteristics of the image，so as to further strengthen
the distinction between the cerebral infarction focus and the normal tissue． Finally，the fuzzy speed function was incor-porated into the active contour model to construct the energy function． The evolution of the contour curve was stopped at the fuzzy boundary of cerebral infarction focus，so that the segmentation of icerebral infarction was realized． The experi-ment results show that the proposed model can achieve accurate segmentation of icerebral infarction focus．

Key words: cerebral infarction, diffusion weighted imaging, Bayesian probability, fuzzy clustering, active con-tour model