Abstract : Alzheimer’s disease (AD) is a progressive neurode- generative condition that features loss of brain volume, loss of neurons, and deposition of amyloid and tau proteins resulting in dementia and cognitive decline. Since AD evolves 8–15 years before the development of overt symptoms and is not treatable, timely diagnosis is imperative to slow down the progression of the disease and enhance patient outcomes. The objective of this research is to classify Alzheimer’s disease stages using deep learning on the OASIS dataset. The model categorizes subjects into four groups: Non-Demented, Very Mild Demented, Mild Demented, and Moderate Demented. Transfer learning is employed, and ResNet-50 as a feature extraction base model and a custom CNN as a classifier. The input has major neuroimaging biomarkers that include hippocampal volume, cortical thickness, and ventricle volume, which improve the model’s diagnostic capability. The use of pre-trained feature extraction with task- specific fine-tuning yields precise and efficient classification by the proposed method. The project assists in early diagnosis and accurate staging of Alzheimer’s disease, facilitating improved treatment and management practices.