Abstract—Brain tumors present a significant challenge in healthcare, necessitating prompt and accurate detection for effective treatment. Pre-trained models were utilized to classify brain tumors without segmentation. Traditional pre-trained architectures like VGG16, VGG19, and ResNet, despite their accuracy, suffer from slow processing speed which leads to use them impractical for rapid diagnosis. Some of the other pre-trained models like Mobile Net and Efficient Net offer fast processing but overfitting problems occur in the small image dataset. To overcome these challenges a two-branch neural network model has been proposed which is lightweight feature extraction and multi-class classification of brain tumors. The proposed two-branch architecture begins with refining the size of the input images, then extraction of robust features, and concludes with a neural network classifier. The proposed model is also evaluated in the presence of image distortions including Gaussian, Poisson, and Speckle noise to ensure the robustness of the proposed solution. Experimental results state that the proposed model is capable of maintaining high accuracy in tumor classification when compared to the other pre-trained models with limited datasets and noise interferences.
 
Keywords—two-branch feature extraction, Magnetic resonance imaging (MRI), brain tumor, Convolutional Neural Network (CNN), VGG16, efficient net, mobile net, DenseNet21


Copyright © 2024 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.