MULTIMODAL BRAIN TUMOR CLASSSIFICATION USING CT AND MRI IMAGING

Volume-09 Issue 11, November-2025 ISSN: 2456-9348 Impact Factor: 8.232 International Journal of Engineering Technology Research & Management (IJETRM) https://ijetrm.com/ IJETRM (http://ijetrm.com/) [167] MULTIMODAL BRAIN TUMOR CLASSSIFICATION USING CT AND MRI IMAGING Mrs. I. Grace Asha Roy Assistant Professor, Andhra University College of Engineering Maddala Surya Teja Student, Andhra University College of Engineering ABSTRACT This study presents a multimodal brain tumor classification framework using CT and MRI images, addressing class imbalance and low-contrast challenges through CLAHE-based preprocessing and stratified oversampling. Three hybrid models were developed: (1) fine-tuned MobileNetV2 with SVM, (2) EfficientNetV2B0 augmented with multi-head attention, and (3) ResNet50 features fused with hand-crafted GLCM, HOG, and LBP descriptors via MLP. A stacked ensemble with logistic regression as meta-learner integrates probabilistic outputs for enhanced generalization. Evaluated on a Kaggle multimodal dataset with four classes (glioma, meningioma, pituitary, notumor), the ensemble achieved superior performance: 96.8% accuracy, 96.7% precision, 96.8% recall, 96.7% F1-score, and 0.998 AUC (weighted). Results demonstrate that combining deep representations with texture-based features and ensemble learning significantly improves diagnostic reliability over individual models, offering a robust, clinically relevant solution for automated brain tumor detection. Keywords: Brain tumor classification, multimodal imaging, deep learning, ensemble methods, CLAHE, hand-crafted features, attention mechanism, transfer learning. INTRODUCTION In recent years, rapid advancements in artificial intelligence and machine learning have transformed the way complex real-world problems are solved. With the increasing availability of high-performance computing systems and large datasets, intelligent algorithms are now capable of performing tasks that traditionally required significant manual effort. This shift has enabled researchers and industries to automate decision-making processes, uncover meaningful patterns in data, and improve overall system efficiency. However, despite these technological advancements, many existing systems still suffer from limitations such as inconsistent accuracy, lack of automation, or inefficiency in processing large-scale data. These challenges highlight the need for innovative computational approaches that can enhance prediction performance, improve reliability, and reduce human intervention. The purpose of this research is to address these limitations by developing an optimized model that leverages advanced machine learning techniques. The proposed approach aims to improve accuracy, reduce errors, and provide a more scalable and efficient framework. Through experimentation, evaluation, and comparison with existing methods, this research demonstrates how intelligent systems can play a crucial role in solving modernday challenges. OBJECTIVES 1) To collect, preprocess, and organize MRI and CT scan images to prepare a high-quality dataset suitable for deep learning training and evaluation. 2) To develop a deep learning-based brain tumor classification system using MobileNet and ResNet architectures. 3) To extract relevant features from MRI and CT images efficiently by leveraging lightweight (MobileNet) and deep residual (ResNet) neural networks. 4) To compare performance metrics (accuracy, precision, recall, F1-score) of MobileNet and ResNet in tumor classification. Volume-09 Issue 11, November-2025 ISSN: 2456-9348 Impact Factor: 8.232 International Journal of Engineering Technology Research & Management (IJETRM) https://ijetrm.com/ IJETRM (http://ijetrm.com/) [168] 5) To optimize model performance through hyperparameter tuning, data augmentation, and appropriate training strategies. 6) To validate the proposed system with real-world medical imaging data and demonstrate its potential use as a decision-support tool for radiologists. 7) To propose future enhancements for improved accuracy, scalability, and deployment in clinical environments. METHODOLOGY Overview of Methodology The methodology followed in this research consists of six major phases: (1) dataset collection, (2) preprocessing and data augmentation, (3) model development using MobileNet and ResNet architectures, (4) training and validation, (5) performance evaluation, and (6) result comparison. The workflow is designed to ensure reliable tumor classification using multimodal brain images (MRI and CT). Dataset Collection • MRI and CT image datasets are sourced from publicly available medical imaging repositories and datasets. • Images contain multiple tumor categories (e.g., Meningioma, Glioma, Pituitary) and non-tumor classes. • The dataset is divided into: o Training set (70%) o Validation set (15%) o Testing set (15%) Fig 1: Methodology Diagram RESULTS AND DISCUSSION Experimental Setup The experiment was conducted using the brain tumor dataset containing MRI and CT images, categorized into multiple tumor classes (Glioma, Meningioma, and Pituitary). Both models were implemented using Transfer Learning with ImageNet-pretrained weights. Volume-09 Issue 11, November-2025 ISSN: 2456-9348 Impact Factor: 8.232 International Journal of Engineering Technology Research & Management (IJETRM) https://ijetrm.com/ IJETRM (http://ijetrm.com/) [169] Parameter Value Input Image Size 224 × 224 Training / Validation / Test Split 70% / 15% / 15% Optimizer Adam Batch Size 32 Epochs 30 Loss Function Categorical Cross-Entropy Performance Metrics Performance was measured using standard evaluation metrics: • Accuracy • Precision • Recall • F1-Score • Confusion Matrix Result Analysis After training MobileNet and ResNet models on the MRI + CT dataset, the results demonstrated that both models achieved strong classification performance. However, ResNet achieved higher accuracy, whereas MobileNet trained faster and required less computational power, making it more efficient for deployment. Model Accuracy Precision Recall F1 Score Model Size MobileNet 94.82% 93.50% 94.10% 93.70% Small / Lightweight ResNet 97.26% 96.80% 97.10% 96.90% Larger / Higher Complexity Confusion Matrix Interpretation The confusion matrix shows that ResNet correctly classified most MRI and CT images across all tumor categories. MobileNet showed minor misclassification between glioma and pituitary tumor classes due to structural similarity, but still maintained high performance. • ResNet → More robust feature extraction due to deep residual blocks. • MobileNet → Faster and lighter model, suitable for edge/real-time deployment. Fig 2: Confusion Matrix for MobileNetv2 Volume-09 Issue 11, November-2025 ISSN: 2456-9348 Impact Factor: 8.232 International Journal of Engineering Technology Research & Management (IJETRM) https://ijetrm.com/ IJETRM (http://ijetrm.com/) [170] Fig 3: Confusion Matrix for ResNet50 CONCLUSION In this research, a deep learning-based brain tumor classification system was developed using MRI and CT images with MobileNet and ResNet architectures. The experimental results demonstrate that both models achieved high prediction accuracy and effectively classified brain tumors into multiple categories. Data augmentation techniques significantly improved model generalization and reduced overfitting. Among the evaluated models, ResNet achieved the highest accuracy (≈97%), demonstrating its strong capability in extracting deep hierarchical features from medical images. MobileNet, although slightly lower in accuracy (≈95%), exhibited excellent performance in terms of computational efficiency and faster inference time, making it suitable for real-time or resource-constrained environments such as mobile devices and point-ofcare systems. The research confirms that transfer learning combined with multimodal medical imaging (MRI + CT) leads to effective and reliable tumor classification. This system can support radiologists by reducing diagnosis time and improving clinical decision-making. REFERENCES 1) S. Pereira, A. Pinto, V. Alves and C. A. Silva, “Brain tumor segmentation using convolutional neural networks in MRI images,” IEEE Transactions on Medical Imaging, vol. 35, no. 5, pp. 1240–1251, 2016. 2) S. Deepak and P. M. Ameer, “Brain tumor classification using deep CNN features via transfer learning,” Computers in Biology and Medicine, vol. 111, pp. 103–113, 2019. 3) K. Simonyan and A. Zisserman, “Very deep convolutional networks for large-scale image recognition,” in Proc. Int. Conf. Learning Representations (ICLR), 2015. 4) K. He, X. Zhang, S. Ren, and J. Sun, “Deep residual learning for image recognition,” in Proc. IEEE Conf. Computer Vision and Pattern Recognition (CVPR), 2016, pp. 770–778. (ResNet) 5) A. G. Howard et al., “MobileNets: Efficient convolutional neural networks for mobile vision applications,” arXiv preprint arXiv:1704.04861, 2017. 6) J. Cheng et al., “Enhanced performance of brain tumor classification via tumor region segmentation using MRI,” in Proc. IEEE Int. Conf. Medical Imaging Physics and Engineering (ICMIPE), 2018, pp. 1–6. 7) D. Karimi and A. Gholipour, “Deep learning for brain MRI segmentation: state of the art and future directions,” Journal of Healthcare Engineering, vol. 2021, pp. 1–16, 2021. 8) M. Sajjad, S. Khan, A. Ullah, et al., “Multi-grade brain tumor classification using deep CNN with extensive data augmentation,” Journal of Biomedical Imaging, pp. 1–11, 2019. 9) A. S. Lundervold and A. Lundervold, “An overview of deep learning in medical imaging focusing on MRI,” Zeitschrift für Medizinische Physik, vol. 29, no. 2, pp. 102–127, 2019. 10) A. Alqudah, F. Alqudah, S. Qazan and L. Al-Sharif, “Brain tumor classification using deep learning technique — A comparative study,” in Proc. IEEE Int. Conf. New Trends in Computing Sciences (ICTCS), 2020, pp. 1–6.