In the last decade, computer-aided early diagnostics of Alzheimer's Disease\n(AD) and its prodromal form, Mild Cognitive Impairment (MCI), has been the\nsubject of extensive research. Some recent studies have shown promising results\nin the AD and MCI determination using structural and functional Magnetic\nResonance Imaging (sMRI, fMRI), Positron Emission Tomography (PET) and\nDiffusion Tensor Imaging (DTI) modalities. Furthermore, fusion of imaging\nmodalities in a supervised machine learning framework has shown promising\ndirection of research.\n In this paper we first review major trends in automatic classification\nmethods such as feature extraction based methods as well as deep learning\napproaches in medical image analysis applied to the field of Alzheimer's\nDisease diagnostics. Then we propose our own design of a 3D Inception-based\nConvolutional Neural Network (CNN) for Alzheimer's Disease diagnostics. The\nnetwork is designed with an emphasis on the interior resource utilization and\nuses sMRI and DTI modalities fusion on hippocampal ROI. The comparison with the\nconventional AlexNet-based network using data from the Alzheimer's Disease\nNeuroimaging Initiative (ADNI) dataset (http://adni.loni.usc.edu) demonstrates\nsignificantly better performance of the proposed 3D Inception-based CNN.\n