Ensemble Model for Detection and Classification of Brain Tumors

Brain tumors, characterized by abnormal cell growth within the brain, present significant challenges for early detection and accurate classification due to their complex and heterogeneous nature. Manual evaluation of MRI scans, though critical, is often laborintensive and prone to errors, highlighting the need for advanced, automated diagnostic methods. Convolutional Neural Networks (CNNs) have demonstrated strong potential in this domain by effectively learning hierarchical features from raw medical images to detect and classify brain tumors. However, single CNN models often encounter difficulties in handling the variability in tumor appearance, such as differences in shape, size, and intensity. To address these challenges, this research proposes an ensemble model that integrates pretrained CNN architectures-ResNet50, Inception, and Xception. The ensemble technique strengthens the adaptability and accuracy of brain tumor identification and categorization by incorporating the advantages of these distinct models. The proposed model was trained and evaluated using a large dataset of 5,316 MRI images, representing a range of brain tumor types, including gliomas, pituitary, meningiomas, and cases no tumor. With a high classification accuracy of 99.39%, our ensemble model proved to be useful in enhancing the reliability of diagnoses and assisting surgeons in making decisions on the treatment of brain tumors.