AI-Based Alzheimer’s Disease Detection Using Deep Learning

Alzheimer’s disease is a progressive neurological disorder that affects memory, cognition, and daily functioning. Early diagnosis plays a critical role in slowing disease progression, yet traditional diagnostic methods rely heavily on expert interpretation of brain scans, which can be time-consuming and subjective. This project was motivated by the need to explore how artificial intelligence can assist clinicians by identifying subtle patterns in brain MRI images that may indicate early stages of Alzheimer’s disease.

For this purpose, we designed a deep learning solution utilizing a pretrained ResNet-18 model for the classification of MRI images into progressive phases of Alzheimer's Disease. The model was properly trained on the OASIS dataset and optimized using techniques for transfer learning. To increase model explainability, we employed Grad-CAM for analyzing the areas of the brain impacting the model's output, thus attempting to fill the gap between AI results and human knowledge.

This project has taught us the values of data quality, model interpretability, and careful evaluation when using AI models in medical domains. By these experiments, we can clearly see that deep learning models are able to assist with early Alzheimer’s diagnoses, and further research can be developed using larger datasets to further improve reliability.