An Explainable and Optimised Framework for Brain Tumor Detection Using EfficientNet-B4 with attention

Nishan, Ashay (2025) An Explainable and Optimised Framework for Brain Tumor Detection Using EfficientNet-B4 with attention. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (881kB) | Preview
Preview	PDF (Configuration Manual) Download (1MB) | Preview

Abstract

Brain tumor classification aims to identify and differentiate various types of tumors from MRI scans to aid in early diagnosis and treatment. Traditional methods, including manual examination by radiologists and basic machine learning models, often struggle with high inter-observer variability, limited scalability, and low accuracy due to the complex nature and variability in tumor shapes and all. These limitations highlight the need for automated, reliable, and interpretable systems. In this study some models used CNN, VGG16, EfficientNet-B4, and EfficientNet-B4 enhanced with an attention mechanism to classify brain MRI images. Two popular optimization techniques, ADAM and SGD are employed to compare learning performance, where ADAM offers faster convergence with adaptive learning rates, and SGD provides more stable generalization. Both optimizers were tested across all models to analyze their impact. Experimental results demonstrate that the EfficientNet-B4 model with attention and the ADAM optimizer achieves the best classification performance. Additionally, Explainable AI using LIME is integrated to highlight important image regions influencing predictions, making the model more transparent and suitable for clinical use.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Yaqoob, Abid UNSPECIFIED
Uncontrolled Keywords:	Brain Tumor Classification; MRI Imaging; Explainable AI (LIME); EfficientNet-B4
Subjects:	R Medicine > RC Internal medicine > RC0254 Neoplasms. Tumors. Oncology (including Cancer) Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence Q Science > Life sciences > Medical sciences > Pathology > Tumors
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	26 Aug 2026 10:06
Last Modified:	26 Aug 2026 10:06
URI:	https://norma.ncirl.ie/id/eprint/9652

Actions (login required)

View Item