Deep Learning-based classification of Alzheimer’s stages: A multiclass approach using MRI data

Alzheimer’s disease (AD), a neurodegenerative disorder that affects millions across the world annually, is recognized by a progressive decline in cognitive functions such as memory, orientation, and reasoning. Despite large advances in the understanding of its pathology, ranging from recent identification of amyloid-β plaques to tau tangles, treatment and early diagnosis remain very elusive. This study presents an enhanced Convolutional Neural Network (CNN) model designed to classify MRI images into four stages of Alzheimer’s disease: non-demented, very mildly demented, mildly demented, and moderately demented. The model incorporates four convolutional layers with ReLU activation, batch normalization, and max-pooling, followed by fully connected layers with dropout regularization to prevent overfitting. Trained on a weighted dataset of 6400 MRI images, the model achieved a peak training accuracy of 99.7% with a final testing accuracy of 88.79% on unseen data. This study ultimately underlines the potential that CNNs hold for early detection and accurate classification of Alzheimer’s disease as a powerful tool for enhancement in diagnostic precision within clinical settings.