Ensemble Deep Learning Models for Enhanced Brain Tumor Classification by Leveraging ResNet50 and EfficientNet-B7 on High-Resolution MRI Images

Abstract

Brain tumors are among the most significant and challenging medical conditions because of their sometimes-fatal outcomes and complexity of treatment options. Early, exact detection of brain tumors is essential for both the most effective treatment programs and much higher patient survival rates. Especially in complex cases, the use of ensemble models has evolved into a quite successful approach to raising diagnostic precision. Combining the features of many deep learning architectures allows ensemble models to improve accuracy and robustness in medical image classification applications. This paper introduces a novel ensemble model using ResNet50 and EfficientNet-B7 to better identify brain tumors using the Crystal Clean: Brain Tumor MRI Dataset. Training over 22,000 high-resolution MRI images in four classes—no tumor, glioma, meningioma, and pituitary tumor—then evaluated models. Validation accuracy of ResNet50 was 97.65%; EfficientNet-B7’s was 98.20%; the ensemble model exceeded single models with a validation accuracy of 99.68%. The post-training analysis verified even more the strength of the ensemble model: an accuracy of 99.53% against 97.4% for ResNet50 and 98.20% for EfficientNet-B7. These results underline the fundamental benefits of using ensemble learning in medical imaging, especially for the brain tumor classification. As it shows better accuracy and dependability, the suggested ensemble model implies that more consistent and reliable brain tumor detection is feasible. This paper focuses on the major contribution modern deep learning approaches provide to improve diagnostic results and offers sharp analysis on the development of artificial intelligence-driven medical imaging.

Keywords

• Brain tumor
• deep learning
• EfficientNet-B7
• ensemble model
• ResNet-50