MRI-based Brain Tumour Classification Using EfficientNetB7 model with transfer learning

Abstract

The brain is the most vital organ in the human body, as it controls and governs all of the body's essential functions. A brain tumour is a bunch of malignant cells that have developed inside the brain. No primary cause has been identified for the formation of tumours in the brain till date. The mortality rate of malignant brain tumours is very high due to the fact that the tumour formation is in the most critical organ of the body. Hence, it is of utmost importance to accurately detect brain tumours at early stages to lower the mortality rate. Although there are many different types of brain tumours, our research will focus on meningioma, pituitary, and glioma tumours. Many efforts are being made to establish a highly accurate solution for the automated classification and segmentation of brain tumours, and Deep Learning algorithms and Machine Learning approaches are playing an immense role in early identification of brain tumours. To beat these challenges, the researchers propose using a variety of pre-trained CNNs to extract features, classify, and segment tumours. We used EfficientNetB7, a pre-trained deep convolutional neural network, to classify brain tumours with an accuracy of 98.4 percent in this study.