Abstract: Brain tumors are among the most challenging health issues because of their complexity and the critical need for early and accurate diagnosis. Magnetic resonance imaging provides excellent spatial resolution and soft tissue contrast, making it an indispensable tool for identifying abnormalities in the brain.
It is highly used for tumor detection in the brain, but MRI scan analysis is time-consuming and prone to human errors because the appearance of a tumor may vary. This paper discusses multiclass classification for brain tumors using CNNs. The work proposes a method with a CNN model implemented in PyTorch to classify the MRI images into four categories: glioma, meningioma, pituitary, and no tumors. The experimental research was conducted using a dataset with varying tumor sizes, locations, shapes, and intensity of the images. For the rigorous evaluation of the model, the dataset of MRI scans was split into the training set and validation set. Techniques like dropout for regularization and data augmentation were used for optimizing CNN architecture to overcome overfitting. Experimental results show that the proposed model has a classification accuracy of 91.4%, which is more accurate than baseline methods. This indicates that it can be efficiently used for brain tumor diagnosis. Results obtained here highlight the potential of deep learning in clinical applications, where the technique provides enhanced diagnostic accuracy and reliability.
Keywords: Convolutional neural networks (CNNs), VGG, Brain Tumors, MRI, Image Classification, Medical Imaging.
|
DOI:
10.17148/IJARCCE.2025.14157
