Implementation of Latest Deep Learning Techniques for Brain Tumor Identification from MRI Images

Abstract

Brain tumors are twisted tissues in which cells repeat rapidly and indefinitely, being unable to control tumor growth. Deep learning uses Magnetic Resonance Imaging (MRI) to detect brain tumors as a method to easily identify brain tumors. A major challenge in identifying brain tumors is the isolation of abnormal cells. This research aims to use Convolutional Neural Networks (CNN), ResNet50 and VGG16 to predict whether a person has a brain tumor from a dataset containing brain MRI images and compare their performance. These methods facilitate the easy detection of tumors for specialists. The accuracy results are stunning and gave a great insight into the behavioural promise of deep learning architectures on specific datasets. The brain MRI images of 3000 patients, of whom 1500 MRI images were tumorous and 1500 images were non-tumorous, were examined. This shows that the deep learning models can greatly contribute to the diagnosis of brain tumours, improving the accuracy and speed of diagnosis. Hence the usage of these models increases for the diagnosis of other complex diseases in the future.