Brain Tumor Detection and Multi-Class Classification Using a Lightweight Convolutional Neural Network

Abstract

The brain tumor is one of the most severe neurological conditions that need timely and proper diagnosis to treat it. The interpretation of Magnetic Resonance Imaging (MRI) scan is usually time-consuming, and it may be varied. This paper introduces a light deep Learning method of automated classification of brain tumors based on the use of a Convolutional Neural Network (CNN). The proposed model classifies MRI images into four classes, namely glioma, meningioma, pituitary tumor, and no tumor. This study preprocessed around 3,000 MRI images in one publicly available dataset by resizing, normalizing, and data augmenting to enhance generalization. The CNN architecture has been optimized to be less complex, having close to 1.2 million trainable parameters, with high performance but being computationally efficient. The model was trained with Adam and categorical cross-entropy loss in 50 epochs. The results obtained through experiments show a total accuracy of 95.2, high precision, recall, and F1-score, which means that the classification was balanced among all types of tumors. As can be seen by comparing the results, the proposed lightweight model can perform with similar results as more complex models like VGG and ResNet with much lower computational cost. The results indicate the possibility of effective deep learning algorithms in clinical decision support in the diagnosis of brain tumors in real-time.

Introduction

The text focuses on the use of deep learning, particularly Convolutional Neural Networks (CNNs), for automated brain tumor classification using MRI images. Brain tumors are serious health conditions requiring accurate and timely diagnosis, and while MRI is the gold standard for detection, manual interpretation is subjective and time-consuming.

Traditional machine learning methods rely on manual feature extraction and segmentation, making them less efficient and sensitive to variations in tumor characteristics. In contrast, CNN-based deep learning models automatically learn features and have significantly improved classification accuracy. However, popular models like VGG and ResNet are computationally expensive, limiting their practical use in real-time clinical environments.

To address this, the study proposes a lightweight CNN architecture that balances high accuracy with low computational complexity. The system is designed to classify MRI images into four categories: glioma, meningioma, pituitary tumor, and no tumor.

The literature review highlights:

• Limitations of traditional methods (manual features, noise sensitivity),

• Effectiveness of deep learning and transfer learning approaches,

• Challenges such as overfitting, high computational cost, and limited datasets,

• CNNs outperforming traditional models like SVM in accuracy and efficiency.

The proposed methodology includes:

1. Data Collection: MRI dataset (~3000 images) from Kaggle with four tumor classes.

2. Preprocessing: Image resizing, normalization, label encoding, and dataset splitting (80:20).

3. Data Augmentation: Techniques like rotation, flipping, and zooming to improve generalization and reduce overfitting.

4. Model Architecture: A lightweight CNN built using Keras with convolutional layers, pooling, dropout, and Global Average Pooling to reduce parameters.

5. Training & Evaluation: The model learns hierarchical features and outputs class probabilities using a SoftMax layer.

The proposed model reduces computational complexity while maintaining competitive accuracy, making it suitable for real-time and resource-constrained clinical applications.

References

[1] Sung, H., Ferlay, J., Siegel, R. L., Laversanne, M., Soerjomataram, I., Jemal, A., & Bray, F. (2021). Global cancer statistics 2020: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA: a cancer journal for clinicians, 71(3), 209-249. [2] Jordan, J. T., & Gerstner, E. R. (2023). Imaging of brain tumors. CONTINUUM: Lifelong Learning in Neurology, 29(1), 171-193. [3] Bauer, S., Wiest, R., Nolte, L. P., & Reyes, M. (2013). A survey of MRI-based medical image analysis for brain tumor studies. Physics in medicine and biology, 58(13), R97-R129. [4] Gordillo, N., Montseny, E., & Sobrevilla, P. (2013). State of the art survey on MRI brain tumor segmentation. Magnetic resonance imaging, 31(8), 1426-1438. [5] LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. nature, 521(7553), 436-444. [6] Swati, Z. N. K., Zhao, Q., Kabir, M., Ali, F., Ali, Z., Ahmed, S., & Lu, J. (2019). Brain tumor classification for MR images using transfer learning and fine-tuning. Computerized Medical Imaging and Graphics, 75, 34-46. [7] Sajjad, M., Khan, S., Muhammad, K., Wu, W., Ullah, A., & Baik, S. W. (2019). Multi-grade brain tumor classification using deep CNN with extensive data augmentation. Journal of computational science, 30, 174-182. [8] Khan MA, Hussain MZ, Mehmood S, Khan MF, Ahmad M, Mazhar T, Shahzad T, Saeed MM. Transfer learning for accurate brain tumor classification in MRI: a step forward in medical diagnostics. Discov Oncol. 2025 Jun 9;16(1):1040. doi: 10.1007/s12672-025-02671-4. PMID: 40490586; PMCID: PMC12149374. [9] Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2012). Imagenet classification with deep convolutional neural networks. Advances in neural information processing systems, 25. [10] [10] D. M. Barode, R. S. Awhad, V. D. Dhangar, and S. S. Kawathekar, “Comparative Analysis of SVM and CNN Techniques for Brain Tumor Detection,” International Journal of Innovative Technology and Exploring Engineering (IJITEE), vol. 13, no. 7, pp. 27–33, Jun. 2024, doi: 10.35940/ijitee.G9908.13070624.

Copyright

Copyright © 2026 Batool Shabbir Hussain, Vaishnavi Kathar, Sarika Thorat, Nandini Kedare, Kalyani Salunke, Dinesh Barode, Seema Kawatheakar. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.