Automatic Brain Tumor Classification Using Transfer Learning

The critical conditions of medical status in humans are neural failures. Out of most destructive ailments, brain tumor is the worst of the kind and reduces hope on life. Inappropriate medical interventions reduce chances of life, particularly exhibiting lesions in large. Therefore, particular and accurate diagnosis treatment is in demand for the patients suffering with brain tumor. Computerization and automation of diagnosis helps an expert to solve the constraints of timeliness, accuracy, and narrowing the point of tumor site. Deep learning has leaped into medical and healthcare systems employing convolution neural networks, transfer learning, RNN, and LSTM. Deep learning classifies brain tumor into meningioma, glioma, and pituitary. A Figshare dataset is tested, augmented, and tested for fitness. A VGG16 architecture is experimented and observed that 98.79% of accuracy is achieved for classification and detection.