Ensemble-Based Hybrid Model for Accurate Brain Tumor Classification Using MR Images

The classification of brain tumors presents a formidable challenge for radiologists, given the intricate features of tumor cells, which can be life-threatening. Nevertheless, the field has witnessed significant progress with the emergence of automated diagnosis techniques that provide valuable support in the detection of brain tumors using MRI. The objectives of the study to tackle a three-class classification problem by distinguishing between glioma, meningioma, and pituitary tumours, all of which are common kinds of brain tumours. To accomplish this, the classification system proposed in this paper employs deep transfer learning, leveraging a pre-trained model such as DenseNet201 and ResNet50V2 to extract pertinent features from brain MRI. The proposed model achieved accuracies of 87.42% and 95.95% with ResNet50V2 and DenseNet201, respectively. However, by implementing the recommended ensemble technique, the overall performance significantly improved, reaching a high accuracy of 96.32 %. Based on our experimental results demonstrate an impressive classification accuracy of up to 96.32 %, surpassing all other existing state-of-the-art methods.