EARLY CLASSIFICATION AND IDENTIFICATION OF BRAIN TUMOR USING DEEP LEARNING TECHNIQUES FROM MRI IMAGES

Abstract

Abstract—The goal of AI is to design machines that can perform tasks as effectively as a human being. Computer tasks involving AI encompass more than just pattern detection, planning, and problem solving. "Deep learning" refers to a set of algorithms used in machine learning. Using data from MRI scans, deep learning models can be developed to aid in the diagnosis and classification of brain tumours. This facilitates the straightforward diagnosis of brain tumours. Most neurological diseases originate from abnormal growth of brain cells, which can compromise brain architecture and even lead to malignant brain tumours. Brain tumour mortality rates could be reduced with better screening and earlier diagnosis. To quickly and accurately spot tumours in MR images, we recommend the convolutional neural network (CNN) based pre-trained EfficientNetB0, EfficientNetB4, and Hybrid transfer learning models proposed here. Recall, loss, accuracy, and AUC were only few of the metrics we used to evaluate the models' efficacy. By comparing the performance of other models to our suggested method utilizing these criteria, I found that a proposed model was superior. We evaluated the suggested models on a dataset of 3264 MR images and found that they achieved impressive results: an accuracy of 98.7%, an AUC of 99.25%, precision of 96%, f1-score of 96%, recall of 99.31%, and a loss of 0.13. We may conclude that the proposed model is useful for early detection of various forms of brain tumours by comparing it to the other models.