Multiclass Brain Tumor Recognition Using Convolutional Neural Network

Abstract

The diagnosis of brain tumors is a tough endeavor that can benefit from computer vision techniques. This study examines the performance of four multiclass brain tumor identification algorithms utilizing magnetic resonance imaging (MRI) data: Convolutional Neural Network (CNN), VGG-16, MobileNetV2, and InceptionV3. The dataset comprises 3264 images of four types of brain cancers (glioma, meningioma, pituitary, and no tumor). The images are pre-processed and then analyzed by the algorithms. The results demonstrate that CNN obtains the highest accuracy of 95%, followed by VGG-16 at 93%, MobileNetV2 at 91%, and InceptionV3 at 89%. This study illustrates the efficiency of CNN in detecting brain cancers and sets a benchmark for future research on the subject.