Volume - 6 | Issue - 3 | september 2024
Published
22 August, 2024
Precise brain tumor classification is essential for efficient diagnosis and treatment planning in the field of medical image analysis. This study investigates hybrid models integrating transfer learning with ensemble methods to enhance classification accuracy. Specifically, the combinations of EfficientNetB3 and VGG19 as feature extractors coupled with Random Forest classifiers. The findings demonstrate significant performance improvements over standalone deep learning approaches. The EfficientNetB3 + Random Forest ensemble achieves an accuracy of 89%, while the VGG19 + Random Forest ensemble achieves 93%, outperforming the KNN+SVM hybrid model. These results highlight the efficacy of using transfer learning for feature extraction and ensemble methods for decision fusion in medical image classification tasks. Moreover, the study contributes insights into optimizing model performance through hyperparameter tuning and data augmentation, essential for enhancing robustness and generalizability across diverse MRI datasets. This research advances the understanding and application of hybrid models in medical imaging, with implications for improving diagnostic accuracy and clinical decision-making.
KeywordsRandom Forest Classifier VGG19 EfficientNetB3 Ensemble Learning Transfer Learning
