Abstarct: Alzheimer's disease is a progressive neurodegenerative disease that causes cognitive impairment and memory loss. Although there is no known cure for Alzheimer's disease but, early detection of the disease can play a serious role in its control and progression. One way to diagnose Alzheimer's disease is to evaluate the operation of different area of the brain using functional magnetic resonance imaging (fMRI). By analyzing the data obtained from the functional magnetic resonance imaging method, the function of each area of the brain can be measured. In this dissertation, the analysis of fMRI data using a convolution neural network-baxsed approach to diagnose Alzheimer's disease is presented. For this purpose, fMRI images in three different groups of healthy people, Alz, and people with Early mild cognitive impairment were collected from the ADNI databaxse. After collecting and pre-processing the fMRI data, a 2-D convolutional neural network model is designed to separate and classify the two classes and is learned on the image datasets. Finally, the accuracy of the proposed model for separating people with early mild cognitive impairment from healthy control and people with Alzheimer's disease is reported to be 74.3 and 78.78, respectively.