Citation :

S. Gayathri, Santhi Baskaran, "Transformer-based Hybrid UCTTransNet–EfficientNet– PVT Framework with NAdam for Brain Tumor Detection and Segmentation," International Journal of Computer Science and Engineering, vol. 13, no. 6, pp. 1-11, 2026. Crossref, https://doi.org/10.14445/23488387/IJCSE-V13I6P101

Abstract

The detection and segmentation of brain tumors are crucial but difficult to achieve because of the heterogeneity of tumors and complicated MRI images. This paper proposes a hybrid framework that combines a transformer with UCTTransNet, EfficientNet, and the Pyramid Vision Transformer (PViT), along with NAdam optimization, to enhance performance. Existing CNN models lack global context, whereas transformer models require high computational cost, which is inefficient and reduces accuracy. The suggested approach involves using EfficientNet to learn strong spatial features and PViT to learn multi-scale contextual dependencies via attention mechanisms. UCTTransNet is more efficient for segmentation because it combines convolutional and transformer encoders, which ensure accurate boundary identification. NAdam optimization speeds up convergence and stabilization of training. Normalization and augmentation are also preprocessing methods that enhance model robustness. The goal is to create a high-performance, computationally efficient framework that improves detection accuracy and segmentation quality while addressing the limitations of existing methods. Experimental results show higher accuracy, a higher Dice coefficient, and better generalization. The hybrid combination greatly enhances the precision of feature representation and segmentation, which is why such a model can be used in automated clinical diagnosis and decision support systems.

Keywords

Brain Tumor Detection, Brain Tumor Segmentation, Transformer-Based Hybrid Framework, Medical Image Analysis, NAdam Optimizer, Attention Mechanism, Feature Extraction, Semantic Segmentation, Deep Learning in Healthcare.

References

1. Manoj Kumar Tyagi et al., Hybrid Deep-Cuckoo Framework for Robust Brain Tumor Detection and Segmentation in MRI Scans, Smart Technologies and Intelligent Computing, 1^st ed., CRC Press, pp. 1-7, 2026.
   [CrossRef]  [Google Scholar] [Publisher Link]
2. Hira Yousaf et al., “Advanced CNN-Based Brain Tumor Detection and Segmentation Using MATLAB: A Diagnostic Accuracy Study,” Pakistan Journal of Medical & Cardiological Review, vol. 5, no. 1, pp. 535-549, 2026.
   [CrossRef] [Google Scholar] [Publisher Link]
3. Ponlatha Sambandham et al., “Brain Tumor Detection using HyGSNet and Feature Extraction with DWT-based GDP,” Journal of Neuroimmunology, vol. 413, 2026.
   [CrossRef] [Google Scholar] [Publisher Link]
4. Homayoun Safarpour et al., “Explainable Deep Learning Framework for Brain Tumor Segmentation using Vision Transformer and Conditional Random Fields,” Multimedia Systems, vol. 32, pp. 1-46, 2026.
   [CrossRef] [Google Scholar] [Publisher Link]
5. Evgin Goceri, and Yuzi D. Winter, “CFATrans: Brain Tumor Segmentation from MRIs using Consecutive Fusion-Attention Transformer with Convolutional Networks and a Composite Loss Function,” Biomedical Signal Processing and Control, vol. 112, 2026.
   [CrossRef] [Google Scholar] [Publisher Link]
6. A. Srinivasa Reddy et al., “T-GAN: Transformer Generative Adversarial Network for Brain Tumor Segmentation,” International Journal of Pattern Recognition and Artificial Intelligence, vol. 40, no. 3, 2026.
   [CrossRef] [Google Scholar] [Publisher Link]
7. Arshleen Kaur et al., “BrainDx: A Dual-Transformer Framework using PVT and SegFormer for Tumor Diagnosis,” Biomedical Signal Processing and Control, vol. 113, pp. 1-19, 2026.
   [CrossRef] [Google Scholar] [Publisher Link]
8. Shakhnoza Muksimova, Jushkin Baltaev, and Young Im Cho, “Brain Tumor Segmentation with Contextual Transformer-Based U-Net,” Electronics, vol. 15, no. 4, pp. 1-17, 2026.
   [CrossRef] [Google Scholar] [Publisher Link]
9. Ameer Hamza, and Robertas Damaševičius, “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, 2025.
   [CrossRef] [Google Scholar] [Publisher Link]
10. Wessam M. Salama, and Moustafa H. Aly, “Brain Tumor Segmentation and Classification: A CVAE-UNETR-ResNet50-VGG16 Hybrid Deep Learning Approach,” Alexandria Engineering Journal, vol. 135, pp. 433-449, 2026.
    [CrossRef] [Google Scholar] [Publisher Link]
11. Yuhui Liao et al., “Trans-MT: A 3D Semi-Supervised Glioma Segmentation Model Integrating Transformer Architecture and Asymmetric Data Augmentation,” Displays, vol. 93, 2026.
    [CrossRef] [Google Scholar] [Publisher Link]
12. G. M. Sasikala, and K. Anand, “ExU-Trans: A Self-Explanatory Transformer with U-Net based Hybrid Model for Brain Tumor Segmentation using MR Imaging,” Complex & Intelligent Systems, vol. 12, pp. 1-25, 2026.
    [CrossRef] [Google Scholar] [Publisher Link]
13. Anita Murmu et al., “SU-FTD2: Transformer-Driven Brain Tumor Imaging Framework Using Explainable AI for Consumer Applications,” IEEE Transactions on Consumer Electronics, vol. 72, no. 2, pp. 4632-4640, 2026.
    [CrossRef] [Google Scholar] [Publisher Link]