A Hybrid Deep Learning-Based Approach for Brain Tumor Classification

Brain tumors (BTs) are spreading very rapidly across the world. Every year, thousands of people die due to deadly brain tumors. Therefore, accurate detection and classification essential in treatment Numerous research techniques have been introduced for BT as well based on traditional machine learning (ML) deep (DL). The ML classifiers require hand-crafted features, which is time-consuming. On contrary, DL robust feature extraction has recently widely used purposes. this work, we propose a hybrid model called DeepTumorNet three types (BTs)—glioma, meningioma, pituitary tumor classification—by adopting basic convolutional neural network (CNN) architecture. GoogLeNet architecture CNN was base. While developing approach, last 5 layers were removed, 15 new added instead these layers. Furthermore, also utilized leaky ReLU activation function map increase expressiveness model. proposed tested publicly available dataset evaluation purposes, it obtained 99.67% accuracy, 99.6% precision, 100% recall, 99.66% F1-score. methodology highest accuracy compared with state-of-the-art results Alex net, Resnet50, darknet53, Shufflenet, GoogLeNet, SqueezeNet, ResNet101, Exception Net, MobileNetv2. showed its superiority over existing models from MRI images.