Abstract:
Alzheimer’s Disease (AD) is a progressive and irreversible neurodegenerative disor-
der. Being the most common cause of dementia, it affects millions of people around
the world, making early detection and diagnosis a necessity. Deep learning can help
detect the numerous patterns associated with this disease, aiding in its early diag-
nosis. In this work, we employ a transfer learning approach to classify MRI images
into Alzheimer’s Disease (AD), Mild Cognitive Impairment (MCI), and Cognitively
Normal (CN) classes by leveraging VGG16 and VGG19 models pre-trained on Im-
ageNet. The datasets used for training are down-sampled and up-sampled datasets
sampled from the ADNI dataset to mitigate the class imbalance issue, resulting in
four experiments. Our approach yielded high accuracy rates ranging from 98.14%
to 99.59%, with VGG19 trained on down-sampled data achieving the highest per-
formance among the four models.
