ABSTRACT Brain tumor becomes a life-threatening disease when it is not identified in its early stages. The manual diagnosis of brain tumors is a complex and time-consuming process. The accurate classification of brain tumor cells is very difficult because of their heterogeneity. In recent years, researchers have carried out many machine learning (ML) and deep learning (DL) based convolutional neural network (CNN) models for the classification of brain tumors to overcome these difficulties. The existing methods must be improved to avoid issues such as slow convergence and poor generalization. The proposed transfer learning model improves the diagnosis and classification of brain tumors and provides accurate outcomes. This helps the doctor give appropriate treatment to the patients. This research proposes an Integrated scaling-based EfficientNetB3V2 fused MaxEnt classifier model to classify brain tumors as pituitary, meningioma, glioma, and no tumor. This model gives excellent results for classification. Investigative results show that the proposed model achieves 99.76% accuracy during training, 99.35% accuracy during validation, and a great classification accuracy of 98.47% during testing compared to other CNN models. This research uses precision, recall, accuracy, F1 score, specificity, true-positive rate, and false-positive rate as performance measures. The existing state-of-the-art CNN models are used for comparative analysis.
Read full abstract