Comparative Analysis of Brain Tumor Classification Using DNN

Brain tumor detection and classification are critical tasks in medical diagnosis, and the use of deep learning models has significantly improved performance in this domain. This paper presents a comparative analysis of eight prominent deep learning architectures—Custom CNN, GoogLeNet, AlexNet, VGG19, DenseNet, CapsNet, U-Net, and ResNet152— applied to the Brain Tumor MRI Dataset from Kaggle. The study evaluates the models, comparing their performances based on accuracy, sensitivity, specificity, among other performance metrics, in order to identify which one is the most effective in classifying MRI brain images into benign, malignant, and normal classes. Although previous works have individually shown the strengths of these models, this paper tries to make a comprehensive comparison in order to analyze their performance under similar conditions. The results show that deep convolutional architectures, such as DenseNet and ResNet, outperform others consistently with regard to classification accuracy, while U-Net performs exceptionally well for tumor segmentation.This study contributes to the growing body of research by highlighting the trade-offs between model complexity, accuracy, and computational requirements, ultimately guiding future research and clinical applications in brain tumor diagnosis.