"Brain Tumour Identification through Advanced Machine Learning Techniqu" by Shaghayegh Khalighiyan

Date of Award

2-19-2025

Publication Type

Master Thesis

Degree Name

M.A.Sc.

Department

Electrical and Computer Engineering

Supervisor

Esam Abdel-Raheem

Supervisor

Mohammad Hassanzadeh

Rights

info:eu-repo/semantics/openAccess

Abstract

Brain tumors are among the most serious and potentially fatal illnesses in the world. If treatment is delayed, their quick growth poses serious hazards and significantly lowers patient survival rates. The large range of symptoms makes diagnosis and therapy even more challenging. Radiologists frequently use magnetic resonance imaging (MRI), which is essential for identifying brain lesions. This study uses convolutional neural network (CNN) architectures to explore two deep learning-based approaches for brain tumor diagnosis and classification. The dataset consists of 6,628 MRI images, including 3,064 tumor images and 3,564 non-tumor images. It was divided into a training set with 5,466 images were classified using an 18-layer CNN and a modified VGG16 model. The model’s 93% accuracy, precision, recall, and F1-score highlight its potential to enhance neuro-oncology diagnostic precision and aid in clinical decision-making. The study used the YOLO (You Only Look Once) framework on two Kaggle open-source datasets (Figshare and Br35H), comprising 5,064 Magnetic Resonance Imaging (MRI) images. It placed particular emphasis on the YOLOv7 algorithm, which has a mean average precision (mAP) of 90.8%, the YOLOv7 model fared somewhat better than the YOLOv5 model, which achieved 90.97%. To enable in-depth study, findings were visually presented and performance measures were carefully computed. These results show how well deep learning methods can advance the identification of brain tumors and offer workable ways to improve neuro-oncological care.

Download

Share

COinS