Comparing different Convolutional Neural Networks for the classification of Alzheimer's Disease

Alzheimer's disease (AD) is a degenerative, incurable neurological disorder that progressively

damages cognitive abilities. AD affects millions of people worldwide. The biopsy method is the

most reliable method to identify AD, but it has the chance to cause irreparable harm. There are

many non-invasive alternatives to biopsies that can be used to diagnose AD without undue risk.

One of these alternatives comprises computer-assisted diagnostic systems, which enable the

identification of brain-impairing illnesses/diseases. This paper used Magnetic Resonance

Imaging scans of brains with four different classes to create a model to detect AD. The opensource

OASIS dataset, which serves as the basis for the study's data, was split into an 80%

training set and a 20% test set. The dataset consisted of classes of moderate AD, mild AD, very

mild AD, and non-AD scans. Five different Convolutional Neural Network methods were used

for classification. The DenseNet-121, ResNet-50, ResNet-18, and AlexNet methods achieved

detection accuracies of 90.5%, 95.1%, 88.4% and 70.5% respectively. The EfficientNetB-7

method failed to identify many cases of AD.