EfficientNetB0’s Hybrid Approach for Brain Tumor Classification from MRI Images Using Deep Learning and Bagging Trees

Brain tumors pose a significant threat to individuals due to their potential to alter DNA and disrupt normal brain function. Early and accurate detection is crucial for successful treatment. In this article, we propose a novel approach for brain tumor classification using a combination of transfer learning and machine learning techniques. The study utilizes two publicly available datasets of MRI images (Figshare and Kaggle) from patients diagnosed with Glioma, Meningioma, and Pituitary tumor. The first approach involves fine-tuning a state-of-the-art convolutional neural network (CNN), EfficientNetB0, for three-class tumor classification. The CNN achieves impressive results with an average accuracy of 97.71% and demonstrates high precision, sensitivity, specificity, and F-measure for all tumor types. In the second approach, high-level features extracted from EfficientNetB0 are fed into a bagging tree classifier. This hybrid approach outperforms the CNN approach, achieving an average accuracy of 99.64%. The bagging tree model shows superior precision, sensitivity, specificity, and F-measure for all tumor types, indicating its effectiveness in classifying tumor severity. The proposed method surpasses existing state-of-the-art models. Accurate and early detection of brain tumors can significantly improve patient prognosis and guide appropriate treatment strategies.