Classification of Magnetic Resonance Imaging (MRI) Scans for Brain Tumor Using Improved EfficientNet Architecture and Transfer Learning

Abstract

Brain tumor is the abnormal growth of cells formed inside the skull. With an average survival rate of 75.7% for all primary brain tumor patients, early detection of brain tumor can significantly reduce the number and severity of cases. The application of deep learning and neural networks in the area of medical diagnosis has macadamized the path of brain tumor detection. Rapid developments have been successfully achieved by the combined application of various medical imaging innovations and deep learning technologies. Convolutional neural networks (CNNs) are the most widely used machine learning algorithm for visual learning and brain tumor recognition. This paper showcases the use of EfficientNet-B7 model and applying fine tuning on hyperparameters in order to come up with a highly accurate model which extracts the features and detects the presence of tumor in brain. The proposed model is successfully able to break down the images and detect the presence of tumor and if the tumor is present, it shows the type of tumor present namely meningioma tumor, pituitary tumor and glioma tumor inside the brain. An accuracy of 98.188% is achieved using our proposed model. The study also highlights a brief comparison of the proposed model with some of the existing model.