Alzheimer's disease (AD) is the most common form of dementia and develops due to the alteration or destruction of brain cells. In the past two decades, the mortality rate associated with AD has increased rapidly. Therefore, this study aims to automate the diagnosis of AD using an ensemble Deep Learning (DL) model and to determine the contribution of Diffusion Tensor Imaging (DTI) images in AD detection, which are effective in showing anatomical structures not visible in Magnetic Resonance Imaging (MRI). The inputs of the 3D Convolutional Neural Network (CNN) developed using residual connections were four different DTI map images: fractional anisotropy, mean diffusivity, radial diffusivity, and axial diffusivity. The outputs of the 3D-CNN model were the AD and healthy control classes. The final output was obtained by combining the results of the four 3D-CNN models, which were trained and tested separately with DTI maps obtained from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. The performance results demonstrated that DTI maps can be used for AD diagnosis. Thus, the intensive labor requirements and human-induced errors inherent in non-automated or semi-automated systems for feature selection have been mitigated, and the developed system can be utilized for AD diagnosis in clinical settings.