The document describes a new computational method for diagnosing Alzheimer's disease (AD) using 3D brain magnetic resonance imaging (MRI) scans. The method involves two phases: 1) segmentation of brain tissues (white matter, grey matter, cerebrospinal fluid) using a convolutional neural network model with Gaussian mixture model input, and 2) classification of AD vs normal controls using a model that combines extreme gradient boosting and support vector machines. The method is evaluated on two datasets, achieving Dice scores of 0.96 for segmentation and accuracies of 0.88 and 0.80 for classification.