Authors - Krish Mithaiwala, Vidhiben Ka Patel, Sachin Patel, Dhwanil Chauhan, Margi Shah, Ankur Patel Abstract - Alzheimer's disease (AD) is a neurological illness that affects memory and cognitive function over time. There is currently no cure or treatment to stop the disease's progression. Early diagnosis is essential but challenging because symptoms typically don't show up until serious brain damage has occurred. This study investigates how convolutional neural networks (CNNs), a type of deep learning, can be used to increase the precision of brain imaging diagnosis. Conv2D, MaxPooling2D, and dense layers were used to train the proposed CNN model on a Kaggle MRI dataset that was divided into three dementia levels: low, mild, and severe. SMOTE was used to resolve class imbalance, improving the model's capacity to categorise under-represented groups. The CNN had a relatively low F1 score, a robust recall, and a high classification accuracy of 96.35%. On the same MRI dataset, comparative comparison with other neural architectures such as VGG16 and Inception V3 also showed dependable performance, underscoring CNNs' potential to support early detection. This study highlights the potential of deep learning to transform Alzheimer's diagnosis and improve clinical decision-making for improved disease treatment, despite difficulties in identifying small brain alterations.
