Deep Learning Study for Image Classification of Alzheimer's MRI
Authors
Castelli Simone
Issue Date
2025
Degree
Data Analytics
Publisher
Dublin Business School
Rights holder
Rights
Abstract
Alzheimer's disease is a leading cause of dementia, impacting millions worldwide. Early diagnosis is crucial for improving patient outcomes, yet traditional methods often lack the sensitivity to detect early-stage neurodegeneration. This study investigates the application of deep learning models for the classification of Alzheimer’s disease using MRI scans. The research follows the CRISP-DM framework and utilises the "Alzheimer MRI Disease Classification Dataset" from Kaggle, comprising 5,120 MRI images categorised into four classes. To enhance classification performance, the dataset was transformed into a binary classification problem—distinguishing between "No Alzheimer" and "Early Alzheimer" cases—while addressing class imbalance through oversampling techniques. Five deep learning architectures were implemented and compared: Convolutional Neural Networks (CNN), Artificial Neural Networks (ANN), Residual Networks (ResNet50), Visual Geometry Group Network (VGGNet16), and Vision Transformers (ViT). The models were trained and evaluated based on accuracy, precision, recall, and F1-score. Results indicate that CNN achieved the highest accuracy (93.46%), followed by ANN (90.10%) and ViT (86.23%), demonstrating their effectiveness in automated MRI-based Alzheimer’s detection. Future work includes refining model interpretability through explainable AI techniques and integrating larger datasets for improved generalisation. This research highlights the potential of deep learning in advancing early Alzheimer’s diagnosis and supporting clinical decision- making.