Alzheimer Detection using Different Deep Learning Methods with MRI Images

Alzheimer is one of the most common diseases that happens for older people. It is a brain disorder that damages human memory. This disease has several levels starting from forgetting a few things up to destroying the ability of doing the simplest tasks. Diagnosing this disease is always a challenging task, especially in the initial stages. In this work, several methods were proposed to detect four different classes of Alzheimer through Brain Magnetic Resonance Imaging (MRI) images. The classes are Mild Demented, Moderate Demented, Non-Demented, and Very Mild Demented. Despite the lack of datasets, a public augmented dataset which is available online for free is used and different methods. Deep learning custom convolution neural network using holdout was used. Additionally, some pre-trained models such as ResNet50 and VGG16 were used. Results were compared and discussed in terms of precision, recall and f-score. Additionally, overfitting was monitored and discussed in each training model. Our work achieved remarkably high and promising results that were more than 90% accuracy and more than 90% F1 score for all classes.