Alzheimer's Disease: Preemptive Diagnosis With Deep Learning Models

This article discusses how to create a classification system that can classify the various stages of Alzheimer's disease. Alzheimer's disease is a mental illness which causes brain shrinkage and brain death. It decreases memory, behavior, and intelligence power. There is no cure for this disease, but medications will improve over time and slow down the symptoms, hence making it vital to detect it early. This study exhibits the usefulness of deep learning methods for early detection of AD. Transfer learning and deep learning techniques are used to build the model. The model is accurate and robust. Our models are CNN, EfficientNetB2, and ResNet50. EfficientNetB2 and ResNet50 are pre-trained models used as the base model with some additional CNN layers. The model is trained using Kaggle's dataset which contains 6400 images. The data was divided into four groups or features: non dementia, mild dementia, very mild dementia, and moderate dementia. We used pre-trained learning models such as EfficientNetb2, ResNet50, and custom CNNs. The test showed that the accuracy of Resnet50 reaches 98%, the accuracy of EfficientNetb2 reaches 99%, and the accuracy of CNN reaches 91%.