Comparison of deep learning methods in brain tumor diagnosis: High-performance classification with MRI data

Year 2025, Volume: 67 Issue: 1, 59 - 73

Banu Ulu

https://doi.org/10.33769/aupse.1619837

Abstract

Brain tumors are serious health problems that must be diagnosed accurately and in a timely manner in order to provide effective treatment. Magnetic resonance imaging (MRI) is widely used in the detection of brain tumors. The accuracy of MRI results depends on the expertise of the physician and usually requires confirmation with biopsy. In recent years, revolutionary developments in image processing and deep learning technologies have provided significant improvements in the diagnosis and classification of brain tumors using MRI. In this study, it is aimed to classify brain tumors accurately and effectively for four different classes (glioma, meningioma, pituitary, and no tumor) previously created using MRI image data. Four different transfer learning-based deep learning methods for classification; ResNet-18, EfficientNet-B0, DenseNet-121, and ConvNeXt-Tiny, are compared using the Fastai library. Accurate diagnosis of brain tumors is of critical importance in the treatment of patients, and the aim of the study is to achieve high accuracy and speed. Our proposed Fastai library-based EfficientNet-B0 model has achieved both fast and highly successful results in the diagnosis of brain tumors with a 99% accuracy rate and 73 minutes of training performance. In addition, the DenseNet-121 model has achieved highly successful results with 99% accuracy rates, and the ResNet-18 and ConvNeXt-Tiny models have achieved 98% accuracy rates. Our results provide fast and effective insights into the possible uses of deep learning frameworks in the field of medical imaging. In addition, these results provide significant improvements compared to studies in the literature.

Keywords

Deep learning, brain tumors, Fastai, magnetic resonance imaging

References

• Prasanthi, T. L., Neelima, N., Improvement of brain tumor categorization using deep learning: a comprehensive investigation and comparative analysis, Procedia Comput. Sci., 233 (2024), 703-712, https://doi.org/10.1016/j.procs.2024.03.259.
• Umarani, C. M., Gollagi, S. G., Allagi, S., Sambrekar, K., Ankali, S. B., Advancements in deep learning techniques for brain tumor segmentation: A survey, Inform. Med. Unlocked, 50 (2024), 1-11, https://doi.org/10.1016/j.imu.2024.101576.
• Appiah, R., Pulletikurthi, V., Esquivel-Puentes, H., Cabrera, C., Hasan, N., Dharmarathne, S., Gomez, L., Castillo, L., Brain tumor detection using proper orthogonal decomposition integrated with deep learning networks, Comput. Methods Programs Biomed., 250 (2024), 1-10, https://doi.org/10.1016/j.cmpb.2024.108167.
• Salati, M., Askerzade, İ., Bostancı, G. E., Convolutional neural network models using metaheuristic based feature selection method for intrusion detection, J. Fac. Eng. Archit. Gazi Univ., 40 (2025), 179-188, https://doi.org/10.17341/gazimmfd.1287186.
• Oltu, B., Karaca, B. K., Erdem, H., Özgür, A., A systematic review of transfer learning based approaches for diabetic retinopathy detection, Gazi Univ. J. Sci., 36 (2023), 1140-1157, https://doi.org/10.35378/gujs.1081546.
• Fastai. Making neural nets uncool again. Available at: https://www.fast.ai/. [Accessed September 2024].
• Zaitoon, R., Syed, H., RU-Net2+: a deep learning algorithm for accurate brain tumor segmentation and survival rate prediction, IEEE Access, 11 (2023), 118105-118123, https://doi.org/10.1109/ACCESS.2023.3325294.
• Jabbar, A., Naseem, S., Mahmood, T., Saba, T., Alamri, F. S., Rehman, A., Brain tumor detection and multi-grade segmentation through hybrid caps-VGGNet model, IEEE Access, 11 (2023), 72518-72536, https://doi.org/10.1109/ACCESS.2023.3289224.
• Smitha, P. S., Balaarunesh, G., Sruthi, Nath, C., Sabatini, S. Aminta., Classification of brain tumor using deep learning at early stage, Measurement: Sensors, 35 (2024), 1-7, https://doi.org/10.1016/j.measen.2024.101295.
• Dikande Simo, A. M., Tchagna Kouanou, A., Monthe, V., Kameni Nana, M., Moffo Lonla, B., Introducing a deep learning method for brain tumor classification using MRI data towards better performance, Inform. Med. Unlocked, 44 (2024), 1-24, https://doi.org/10.1016/j.imu.2023.101423.
• Al-Jammas, M., Al-Sabawi, E., Yassin, A., Abdulrazzaq, A., Brain tumors recognition based on deep learning, e-Prime Adv. Electr. Eng. Electron. Energy, 8 (2024), 1-8, https://doi.org/10.1016/j.prime.2024.100500.
• Gürkahraman, K., Karakış, R., Brain tumors classification with deep learning using data augmentation, J. Fac. Eng. Archit. Gazi Univ., 36 (2021), 997-1011, https://doi.org/10.17341/gazimmfd.762056.
• Arı, A., Hanbay, D., Tumor detection in MR images of regional convolutional neural networks, J. Fac. Eng. Archit. Gazi Univ., 34 (2019), 1395-1408, https://doi.org/10.17341/gazimmfd.460535.
• MRI for Brain Tumor with Bounding Boxes, MRI images for brain tumors for object detection or classification. Available at: https://www.kaggle.com/ datasets/ ahmedsorour1/mri-for-brain-tumor-with-bounding-boxes. [Accessed September 2024].