Hybrid Deep Learning MRI Classification Using DenseNet201, EfficientNetB2, and Vision Transformer for Early Detection of Alzheimer

Jelagat, Lauryn (2024) Hybrid Deep Learning MRI Classification Using DenseNet201, EfficientNetB2, and Vision Transformer for Early Detection of Alzheimer. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (609kB) | Preview

Abstract

Alzheimer’s disease is a global crisis in the medical field which can be managed with early detection or diagnosis. However, the change of the brain is often so subtle to identify hence posing a great challenge in the early detection of the disease. This study proposes a hybrid model that integrates DenseNet201, EfficientNetB2, and Vision Transformer (ViT) to enhance MRI classification. DenseNet201 extracts fine-grained spatial details, EfficientNetB2 captures mid-level structural patterns, and ViT models global contextual dependencies, enabling the model to comprehensively analyze MRI images.

The hybrid model was trained and evaluated on a large, imbalanced dataset, achieving an overall accuracy of 95%. It demonstrated high recall (97%) for the underrepresented "Moderate Demented" class, addressing a critical challenge in imbalanced datasets. The weighted F1-score of 0.95 further confirms the model's ability to balance precision and recall across all diagnostic categories. However, the study also highlights limitations, such as the high computational cost and dependency on pre-trained ImageNet weights, which may restrict generalizability to diverse MRI datasets.

This research demonstrates the potential of hybrid deep learning models in advancing the early diagnosis of neurodegenerative diseases and suggests future directions, including optimization for resource-constrained environments and the incorporation of explainability techniques to enhance clinical adoption.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Singh, Jaswinder UNSPECIFIED
Uncontrolled Keywords:	Magnetic Resonance Imaging; Vision Transformer (ViT); EfficientNet; DenseNet
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science R Medicine > RC Internal medicine > RC0321 Neuroscience. Biological psychiatry. Neuropsychiatry R Medicine > Healthcare Industry
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	02 Sep 2025 14:42
Last Modified:	02 Sep 2025 14:42
URI:	https://norma.ncirl.ie/id/eprint/8715

Actions (login required)

View Item