CLASSIFICATION OF BRAIN TUMOUR MRI IMAGES USING ENSEMBLE MODEL

Abstract

Research on Brain Tumours (BT) is increasingly moving toward automated diagnostic tools that can assist clinicians in making quicker and more accurate decisions. Since brain tumours are challenging to diagnose and can be life-threatening, early identification remains critical for improving treatment outcomes. Recent progress in machine learning (ML), deep learning (DL), and MRI-based imaging has helped overcome issues seen in manual interpretation, such as variation in expert judgment and the complex nature of tumour patterns. In this study, several deep learning architectures-including Convolutional Neural Networks (CNNs), ResNet50, VGG-16, and Inception V3-were used to develop an automated detection system capable of producing fast and dependable predictions. By combining MRI scans with features generated by these models, the system reduces manual effort and simplifies the diagnostic process. The study also uses an ensemble strategy, merging outputs from multiple networks to improve accuracy and model stability. When evaluated on 259 MRI images, the ensemble achieved 90% accuracy, supporting existing findings that DL methods are highly effective in tumour classification. Overall, the proposed system shows strong promise for improving automated brain tumour detection and aiding clinical workflows.