Multi-modal MRI-Based Classification of Brain Tumors. A Comprehensive Analysis of 17 Distinct Classes

Brain tumor classification is a critical task in medical imaging, as accurate and timely diagnosis is essential for effective treatment planning and patient care. In this study, we present a comprehensive analysis of brain tumor classification using a multi-modal MRI dataset comprising 17 distinct tumor classes. The dataset is split into training (60%), validation (20%), and testing (20%) sets, ensuring a robust evaluation of our proposed approach. We employ the Xception pre-trained model, a state-of-the-art deep learning architecture, to extract high-level features from multi-modal MRI data. The model is fine-tuned on the training set and evaluated on the validation and testing sets. Our results demonstrate remarkable performance with a training accuracy of 0.9991 and a training loss of 0.0026, reflecting the model’s ability to capture intricate patterns within the data. During validation, the model achieves an accuracy of 0.9791 and a loss of 0.0784, further confirming its effectiveness in classifying brain tumors across various modalities. When evaluated on the testing set, the model achieves a robust accuracy of 0.9788 with a loss of 0.0836, demonstrating its generalization capability to unseen data. Moreover, our evaluation includes essential metrics such as F1-score (0.9788), recall (0.9788), and precision (0.9791), affirming the model’s balanced performance across the 17 distinct tumor classes. Additionally, the Receiver Operating Characteristic (ROC) curve analysis for each class shows excellent discriminative power with an area under the curve (AUC) of 1.000. The proposed approach showcases promising results in accurately classifying brain tumors, highlighting the potential of leveraging deep learning and multi-modal MRI data for improved diagnostic capabilities.