Alzheimer's disease (AlzheimerÂÂs disease, AD) is a common chronic mental decline disease with highly morbidity. Mild cognitive impairment is a transitional stage between normal aging and AD. Currently the analysis of magnetic resonance imaging of these two diseases mainly judged through the doctor's experience with a heavy workload and a strong subjective. To solve this problem, a combination of structural and functional features was proposed in this study based on the multi-modal magnetic resonance imaging data, rather than only relying on a single structure or functional property to distinguish two types of diseases in previous work. In this research, the multi-modal magnetic resonance imaging data were used to construct rest functional networks, then, the network properties and the gray matter volume of atrophy gray matter in structures images were extracted as the classification features to train the SVM classifier. Experimental results show that the combination of structural and functional characteristics that can differentiate MCI and normal control (accuracy of 91.7%), AD and normal controls (accuracy of 100%), AD and MCI (accuracy of 87.8%), effectively improve the classification accuracy rate of two types of diseases. The method can be verified as an auxiliary diagnostic model of Alzheimer's disease in the future.