Deep learning for magnetic resonance imaging brain tumor detection: evaluating ResNet, EfficientNet, and VGG-19

This paper investigates the application of convolutional neural networks (CNNs) for the early detection of brain tumors to enhance diagnostic accuracy. Brain tumors present a significant global health challenge, and early detection is vital for successful treatments and patient outcomes. The study i...

Full description

Bibliographic Details
Published in:International Journal of Electrical and Computer Engineering
Main Author: Muftic F.; Kadunic M.; Musinbegovic A.; Almisreb A.A.; Jaafar H.
Format: Article
Language:English
Published: Institute of Advanced Engineering and Science 2024
Online Access:https://www.scopus.com/inward/record.uri?eid=2-s2.0-85206005418&doi=10.11591%2fijece.v14i6.pp6360-6372&partnerID=40&md5=e6eb00e23e56b7278f05a6ee13fd58d5
Description
Summary:This paper investigates the application of convolutional neural networks (CNNs) for the early detection of brain tumors to enhance diagnostic accuracy. Brain tumors present a significant global health challenge, and early detection is vital for successful treatments and patient outcomes. The study includes a comprehensive literature review of recent advancements in brain tumor detection techniques. The main focus is on the development and evaluation of CNN models, including EfficientNetB3, residual networks-50 (ResNet50) and visual geometry group-19 (VGG-19), for binary image classification using magnetic resonance imaging (MRI) scans. These models demonstrate promising results in terms of accuracy, precision, and recall metrics. However, challenges related to overfitting and limited dataset size are acknowledged. The study highlights the potential of artificial intelligence (AI) in improving brain tumor detection and emphasizes the need for further research and validation in real-world clinical settings. EfficientNetB3 reached 99.44% training accuracy but showed potential overfitting with validation accuracy dropping to 89.47%. ResNet50 steadily improved to 83.62% training accuracy and 89.47% validation accuracy. VGG19 struggled, with only 62% accuracy. © 2024 Institute of Advanced Engineering and Science. All rights reserved.
ISSN:20888708
DOI:10.11591/ijece.v14i6.pp6360-6372