A Hybrid GAN-CNN Framework for Enhanced Brain Tumor Classification from MRI Images

Abstract

This paper introduces a hybrid deep learning model, a combination of Generative Adversarial Networks (GANs) and Convolutional Neural Networks (CNNs) to improve brain tumor classification with MRI images. The diagnosis of brain tumors has been a major dilemma in medical imaging because of the complexity of the tumor structures, lack of data, and imbalance of the classes. To overcome these issues, the proposed framework encourages the use of GAN-based data augmentation in order to create real-world synthetic MRI images, which will enhance the dataset and enhance the generalization of the models. Moreover, traditional preprocessing and augmentation methods like normalization, contrast enhancement, and noise reduction are used to enhance the quality of the data. This is followed by the design of a strong CNN architecture to automatically project hierarchical features as well as multi-classify into glioma, meningioma, pituitary tumor, and no tumor. The model is trained and tested on a rich dataset of publicly accessible sources, making sure that there is diversity in the imaging perspectives and conditions. The experimental results show that the proposed method has a high classification rate of around 99 percent and a low loss value, which reflects high predictive performance and stability. The hybrid GAN-CNN model, compared to the current methods, demonstrates superior performance in comparison to the traditional CNN-based methods, as well as other hybrid techniques. Moreover, the analysis of confusion matrix and ROC curve proves that the model is able to differentiate between the various classes of tumors effectively. The suggested framework provides a credible and scalable solution to automated brain tumor diagnosis, and it has a high probability of its clinical use.