Diffusion kurtosis imaging(DKI)is a simple extension of the diffusion tensor imaging(DTI)technique and widely used in clinic recently.However,the advantages of DKI over DTI are not very clear,especially the characteristic of kurtosis was beyond common understanding.In this paper,SVM-RFE based pattern recognition was performed on pathological changes detecting in Alzheimers compared to the normal healthy.Firstly the DTI-derived and the DKI-derived indexes were calculated and 23 regions of interest(ROIs)were defined.Statistical analysis,pattern recognition were conducted respectively as well as the permutation test,ROC analysis and Pearson correlation.The DKI-derived dataset were revealed to be more effective with better classification performance than the DTI-derived dataset and the kurtosis indexes that describe the diffusion differently also shown high performance as the diffusivity can reach or even better.The results of feature selection were in accordance with the pathological researches.Pattern recognition can be served as diagnostic tool and DKI is a powerful model that can provide a dataset with both of the diffusivity and kurtosis which complement each other to describe the micro-changes in diseases like Alzheimers.