Abstract: According to the characteristic of medical image dataset，this paper introduces a new method of data mining based on rough set theory and decision tree，which applies the discretization method on the basis of the importance of attribute in rough set to discretize the feature and reduce attributes，then gets the low dimention training data，and finally makes use of the SLIQ algorithm of decision tree to extract classification rules.Lots of experiments prove that the new method of the combination of rough set theory and SLIQ algorithm applied in medical image classification improve recogniton rate and efficiency while reduce the irrelevant feature with classification and keep the inherent property of raw data.