Alzheimer's disease (AD) is a neuro-degenerative disease that can cause dementia and result severe reduction in brain function inhibiting simple tasks especially if no preventative care is taken. Over 1 in 9 Americans suffer from AD induced dementia and unpaid care for people with AD related dementia is valued at $271.6 billion. In this paper, we first review other approaches that could be used for early detection of AD. We then give an overview of our dataset that was from the Alzheimer's Disease Neuroimaging Initiative (ADNI) and propose a deep Convolutional Neural Network (CNN) architecture consisting of 7,866,819 parameters. This model has three different length convolutional branches each comprised of different kernel sizes that can predict whether a patient is non-demented, mild-demented, or moderately-demented with a 99.05% three class accuracy.