Multi-Class Brain Tumor Classification in MRI Using Transfer Learning and Explainable AI: A Comparative Study of Grad-CAM, Grad-CAM++, and Score-CAM
Keywords:
Brain Tumor Classification, Magnetic Resonance Imaging (MRI), Transfer Learning, Explainable Artificial Intelligence (XAI), Grad-CAM, Grad-CAM++, Score-CAMAbstract
Brain tumor classification from magnetic resonance imaging (MRI) remains a challenging task due to high intra-class variability and strong inter-class similarity among tumor categories. This study presents a unified deep learning framework for multi class brain tumor classification by integrating transfer learning with multiple explainable artificial intelligence (XAI) techniques. Three convolutional neural network architectures—ResNet-50, EfficientNet-B0, and DenseNet121—were fine-tuned and evaluated using a publicly available four-class MRI dataset consisting of glioma, meningioma, pituitary tumor, and no tumor categories. The dataset was systematically divided into training, validation, and testing subsets under consistent experimental conditions to ensure reliable comparative evaluation. Experimental results demonstrated strong classification performance across all evaluated models, with accuracy values ranging from 96% to 97%. ResNet50 and DenseNet121 exhibited more stable convergence behavior and more consistent interpretability characteristics, while EfficientNet-B0 achieved competitive performance with efficient feature scaling. Receiver Operating Characteristic (ROC) analysis further confirmed the strong discriminative capability of the models, with Area under the Curve (AUC) values approaching 1.0 across multiple classes. To improve model interpretability, a comparative explainability analysis was conducted using Grad-CAM, Grad-CAM++, and Score-CAM. The results revealed substantial variation in localization behavior across explainability methods. Grad-CAM++ provided more concentrated spatial localization, whereas Score-CAM generated smoother and less noisy activation maps. However, the explainability outputs remained primarily qualitative and did not achieve precise tumor boundary delineation, indicating that fine-grained localization and clinically validated interpretability remain open challenges. The principal contribution of this work lies in the development of a unified comparative framework that jointly evaluates classification performance, training stability, and explainability behavior across multiple CNN architectures under controlled experimental settings. Overall, the proposed framework provides a reliable and interpretable comparative framework for explainable brain tumor classification using transfer learning.
References
[1] F. J. Dorfner, J. B. Patel, J. Kalpathy-Cramer, E. R. Gerstner, and C. P. Bridge, “A review of deep learning for brain tumor analysis in mri,” NPJ Precision Oncology, vol. 9, no. 1, p. 2, 2025.
[2] S. Bouhafra and H. El Bahi, “Deep learning approaches for brain tumor detection and classification using mri images (2020 to 2024): A systematic review,” Journal of Imaging Informatics in Medicine, vol. 38, no. 3, pp. 1403–1433, 2025.
[3] A. B. Abdusalomov, M. Mukhiddinov, and T. K. Whangbo, “Brain tumor detection based on deep learning approaches and magnetic resonance imaging,” Cancers, vol. 15, no. 16, p. 4172, 2023.
[4] S. Anantharajan, S. Gunasekaran, T. Subramanian, et al., “Mri brain tumor detection using deep learning and machine learning approaches,” Measurement: Sensors, vol. 31, p. 101026, 2024.
[5] M. S. Ullah, M. A. Khan, A. Masood, O. Mzoughi, O. Saidani, and N. Alturki, “Brain tumor classification from mri scans: a framework
of hybrid deep learning model with bayesian optimization and quantum theory-based marine predator algorithm,” Frontiers in Oncology, vol. 14, p. 1335740, 2024.
[6] E. Albalawi, A. Thakur, D. R. Dorai, S. Bhatia Khan, T. Mahesh, A. Al musharraf, K. Aurangzeb, and M. S. Anwar, “Enhancing brain tumor classification in mri scans with a multi-layer customized convolutional neural network approach,” Frontiers in computational neuroscience, vol. 18, p. 1418546, 2024.
[7] D. Rastogi, P. Johri, M. Donelli, S. Kadry, A. A. Khan, G. Espa, P. Feraco, and J. Kim, “Deep learning-integrated mri brain tumor analysis: feature extraction, segmentation, and survival prediction using replicator and volumetric networks,” Scientific Reports, vol. 15, no. 1, p. 1437, 2025.
[8] A. Hamza and R. Damaˇ seviˇcius, “Deep learning for brain tumor segmentation and classification: a systematic review of methods and trends,” Computers, materials and continua., vol. 86, no. 1, pp. 1–41, 2026.
[9] T. Berghout, “The neural frontier of future medical imaging: A review of deep learning for brain tumor detection,” Journal of imaging, vol. 11, no. 1, p. 2, 2024.
[10] T. A. Fahim, F. B. Alam, and M. A. Hossain, “Brain tumor detection, classification and segmentation by deep learning models from mri images: Recent approaches, challenges and future directions,” Array, p. 100571, 2025.
[11] R. Nimmagadda and P. K. Devi, “A deep learning approach for brain tu mour classification and detection in mri images using yolov7,” Frontiers in Oncology, vol. 15, p. 1508326, 2025.
[12] M. N. E. Farandi, A. K. Muda, S. Winarno, and H. Basiron, “Comparative study of deep learning models for mri-based brain tumor classification,” Journal of Future Artificial Intelligence and Technologies, vol. 2,no. 3, pp. 370–387, 2025.
[13] E. Chukwujindu, H. Faiz, A.-D. Sara, K. Faiz, and A. De Sequeira, “Role of artificial intelligence in brain tumour imaging,” European journal of radiology, vol. 176, p. 111509, 2024.
[14] Z. U. Abidin, R. A. Naqvi, A. Haider, H. S. Kim, D. Jeong, and S. W. Lee, “Recent deep learning-based brain tumor segmentation models using multi-modality magnetic resonance imaging: A prospective survey,” Frontiers in Bioengineering and Biotechnology, vol. 12, p. 1392807, 2024.
[15] C. Reddy, P. A. Reddy, H. Janapati, B. Assiri, M. Shuaib, S. Alam, and A. Sheneamer, “A fine-tuned vision transformer based enhanced multi class brain tumor classification using mri scan imagery,” Frontiers in oncology, vol. 14, p. 1400341, 2024.
[16] G. Jayaraman, S. Meganathan, S. S. M. Shah, M. Anuradha, R. K. Sundararajan, and R. Rajakumar, “A hybrid cnn–vit framework with cross-attention fusion and data augmentation for robust brain tumor classification,” Scientific Reports, 2025.
[17] M. Rasool, A. Noorwali, H. Ghandorh, N. A. Ismail, and W. M. Yafooz, “Brain tumor classification using deep learning: A state-of the-art review,” Engineering, Technology & Applied Science Research, vol. 14, no. 5, pp. 16586–16594, 2024.
[18] W. Bukaita and V. Vadde, “Comparative evaluation of cnn and resnet18 architectures for mri-based brain tumor classification using deep learning,” Medical Research Archives, vol. 13, no. 12, 2025.
