MRI Brain Tumor Classification Using Deep Convolutional Neural Network

Abstract

A brain tumor is one of the most complex disorders that occurs when the brain cells begin to grow uncontrollably. The biggest issue before starting treatment is detecting and classifying tumors from brain magnetic resonance imaging (MRI) scans. For ages, researchers have worked hard to develop the best approach for real-life medical image recognition with greater precision. The current approach is inconvenient, time-consuming, and human error-prone. These flaws emphasize the significance of establishing a fully automated deep learning-based brain tumor classification approach. The purpose of this article is to develop a convolutional neural network (CNN) to classify brain tumors to make an early diagnosis. Firstly, input images are resized and grayscale conversion is applied to these input photos, which aids in the reduction of complexity. After that, data augmentation is applied to enhance the data number, and these augmented images are converted into binary pictures to distinguish an object from its surroundings. The proposed CNN consists of three convolutional 2D layers, three max-pools, two completely interconnected layers, ReLU activation functions, one layer of cross-channel normalization, and a dropout layer. The proposed network structure performs admirably, with an overall accuracy of 96.9%. The results are compared to previous research work and outperform many state-of-the-art approaches.