Adaptive Sparse Cross-Scale Transformer for Computationally Efficient Brain Tumor MRI Segmentation

Ravichandra Bandi; Selvanayaki  S; Amitha Ida  Chandran; Sudha  Subramaniam; Suganthi  R; Anand   Rajendran

doi:10.35882/jeeemi.v8i4.1742

Ravichandra Bandi Department of ECE, N S Raju Institute of Technology (A), Sontyam, Visakhapatnam, India https://orcid.org/0000-0002-5165-4018
Selvanayaki S Department of Artificial Intelligence and Data Science, Saveetha Engineering College, Chennai, India https://orcid.org/0000-0003-2652-8236
Amitha Ida Chandran Department of Computer Science and Engineering, KPR Institute of Engineering and Technology, Coimbatore, India https://orcid.org/0000-0003-0114-8237
Sudha Subramaniam Department of ECE, Kongu Engineering College, Perundurai, India https://orcid.org/0000-0003-0351-1140
Suganthi R Department of Electronics and Communication Engineering, Panimalar Engineering College, Chennai, India https://orcid.org/0000-0002-7045-5321
Anand Rajendran Department of Aeronautical Engineering, Nehru Institute of Technology, Coimbatore, India https://orcid.org/0000-0001-5170-2887

DOI: https://doi.org/10.35882/jeeemi.v8i4.1742

Abstract

Brain tumor segmentation from Magnetic Resonance Imaging (MRI) is a critical task in medical image analysis, owing to the irregular morphology of tumors, heterogeneous intensity distributions across MRI modalities, and the presence of multiple overlapping sub-regions. Accurate delineation of these regions is essential for clinical diagnosis, treatment planning, and longitudinal disease monitoring. Although Vision Transformer (ViT)-based architectures have demonstrated promising performance in medical image segmentation by capturing long-range dependencies and global contextual information, conventional multi-scale transformer models remain constrained by static grouped attention mechanisms that introduce redundant computations and exhibit quadratic complexity with respect to input size. To address these limitations, this paper proposes the Adaptive Sparse Cross-Scale Transformer (ASCT), a computationally efficient framework for brain tumor MRI segmentation. ASCT incorporates three key innovations: (i) a Dynamic Scale Routing (DSR) module that adaptively weights multi-scale features using learned routing coefficients, replacing fixed channel grouping; (ii) a Sparse Token Attention (STA) mechanism that restricts attention computation to the most informative token pairs, reducing complexity from quadratic O(N²) to near-linear O(Nk); and (iii) a linearized attention approximation that significantly reduces GPU memory consumption during training. Additionally, cross-scale feature fusion is performed prior to the attention operation to suppress redundant computations and enhance inter-scale feature interaction. The proposed ASCT model is evaluated on the BraTS 2021 dataset for multi-region segmentation, targeting Whole Tumor (WT), Tumor Core (TC), and Enhancing Tumor (ET) sub-regions. Experimental results demonstrate that ASCT achieves Dice scores of 89.2%, 84.1%, and 83.5% for WT, TC, and ET, respectively, yielding an average Dice score of 85.6%. Compared to the baseline Vision Transformer, the proposed model reduces computational complexity from 94.6 GFLOPs to 58.4 GFLOPs and GPU memory usage from 11.2 GB to 7.3 GB, confirming its efficiency and practical viability for real-world clinical applications.

Downloads

Download data is not yet available.

References

Abidin, Z. U., Naqvi, R. A., Haider, A., Kim, H. S., Jeong, D., and Lee, S. W., “Recent deep learning-based brain tumor segmentation models using multi-modality magnetic resonance imaging: A prospective survey,” Frontiers in Bioengineering and Biotechnology, vol. 12, Art. no. 1392807, 2024, doi: 10.3389/fbioe.2024.1392807

Senthil Kumar, A. V., Kartiga, N., Kaur, G., Musirin, I. B., Jagadamba, G., Vanishree, G., Kumar, R., et al., “Automated segmentation of brain tumor MRI images using machine learning,” in Common Infectious Diseases of the Central Nervous System: Pathologies, Diagnostics, and Therapeutics, pp. 241–264, IGI Global Scientific Publishing, 2026, doi: 10.4018/979-8-3693-4070-7.ch009.

Bhagyalaxmi, K., Dwarakanath, B., and Reddy, P. V. P., “Deep learning for multi-grade brain tumor detection and classification: A prospective survey,” Multimedia Tools and Applications, vol. 83, pp. 65889–65911, 2024, doi: 10.1007/s11042-024-18129-8.

Kumar, S. S., and Vinod Kumar, R. S., “Advances in brain tumor segmentation using transformer models,” in Transforming Healthcare with AI and IoT, pp. 228–252, CRC Press, 2026, doi: 10.1201/9781003658504.

Balasamy, K., Krishnaraj, N., and Vijayalakshmi, K., “An adaptive neuro-fuzzy-based region selection and authenticating medical image through watermarking for secure communication,” Wireless Personal Communications, vol. 122, pp. 2817–2837, 2022, doi: 10.1007/s11277-021-09031-9.

Dataset collection: https://www.kaggle.com/datasets/syedsajid/brats2021

Albalawi, E., Thakur, A., Dorai, D. R., Khan, S. B., Mahesh, T. R., Almusharraf, A., Aurangzeb, K., and Anwar, M. S., “Enhancing brain tumor classification in MRI scans with a multi-layer customized convolutional neural network approach,” Frontiers in Computational Neuroscience, vol. 18, Art. no. 1418546, 2024, doi: 10.3389/fncom.2024.1418546.

Rasool, N., Wani, N. A., Bhat, J. I., Saharan, S., Sharma, V. K., Alsulami, B. S., Alsharif, H., and Lytras, M. D., “CNN-TumorNet: Leveraging explainability in deep learning for precise brain tumor diagnosis on MRI images,” Frontiers in Oncology, vol. 15, Art. no. 1554559, 2025, doi: 10.3389/fonc.2025.1554559.

Semwal, T., Jain, S., Mohanta, A., et al., “A hybrid CNN-SVM model optimized with PSO for accurate and non-invasive brain tumor classification,” Neural Computing and Applications, vol. 37, pp. 27901–27930, 2025, doi: 10.1007/s00521-025-11078-9.

Chandraprabha, K., Ganesan, L., and Baskaran, K., “A novel approach for the detection of brain tumor and its classification via end-to-end vision transformer-CNN architecture,” Frontiers in Oncology, vol. 15, Art. no. 1508451, 2025, doi: 10.3389/fonc.2025.1508451.

Isensee, F., Jaeger, P.F., Kohl, S.A., Petersen, J., Maier-Hein, K.H., “nnU-Net: a selfconfiguring method for deep learning-based biomedical image segmentation”, Nat. Methods vol.18, no.2, 203–211, 2021, doi.org/10.1038/s41592-020-01008-z.

Gupta, S., and Abd Aziz, A., “Advancing brain tumor classification through pre-trained transformer and transfer learning models,” Franklin Open, vol.14, Art. no. 100493, 2026, doi: 10.1016/j.fraope.2026.100493.

Shamia, D., Balasamy, K., and Suganyadevi, S., “A secure framework for medical image by integrating watermarking and encryption through fuzzy-based ROI selection,” Journal of Intelligent & Fuzzy Systems, vol. 44, no. 5, pp. 7449–7457, 2023, doi: 10.3233/JIFS-222618.

Balasamy, K., Seethalakshmi, V., and Suganyadevi, S., “Medical image analysis through deep learning techniques: A comprehensive survey,” Wireless Personal Communications, vol. 137, pp. 1685–1714, 2024, doi: 10.1007/s11277-024-11428-1.

Suganyadevi, S., and Seethalakshmi, V., “Deep recurrent learning-based qualified sequence segment analytical model (QS2AM) for infectious disease detection using CT images,” Evolving Systems, vol. 15, pp. 505–521, 2024, doi: 10.1007/s12530-023-09554-5.

Salama, W. M., and Aly, M. H., “Brain tumor segmentation and classification: A CVAE-UNETR-ResNet50-VGG16 hybrid deep learning approach,” Alexandria Engineering Journal, vol. 135, pp. 433–449, 2026, doi: 10.1016/j.aej.2026.01.003.

Maheshwari, D., Kelkar, T., and Francis, S., “An optimized hybrid deep learning-based approach for brain tumor classification,” Biomedical Signal Processing and Control, vol. 113, Art. no. 108935, 2026, doi: 10.1016/j.bspc.2025.108935.

Gutta, S., and Yathirajam, S. S., “Improving glioma grade classification with a hybrid CNN-transformer model,” Intelligence-Based Medicine, vol. 13, Art. no. 100334, 2026, doi: 10.1016/j.ibmed.2025.100334..

Reddy, A. S., Pemula, R., Raju, K. K., and Murty, C. S. V. V. S. N., “T-GAN: Transformer generative adversarial network for brain tumor segmentation,” International Journal of Pattern Recognition and Artificial Intelligence, vol. 40, no. 3, Art. no. 2557024, 2026, doi: 10.1142/S0218001425570241.

Almuhaimeed, A., Bilal, A., Alzahrani, A., Alrashidi, M., Alghamdi, M., and Sarwar, R., “Brain tumor classification using GAN-augmented data with autoencoders and Swin transformers,” Frontiers in Medicine, vol. 12, Art. no. 1635796, 2025, doi: 10.3389/fmed.2025.1635796.

Guan, Y., Alshammari, A., Wang, Y., Gul, J. Z., and Imran, A., “ResSGA-Net: A deep learning approach for enhanced brain tumor detection and accurate classification in healthcare imaging systems,” Journal of Genetic Engineering and Biotechnology, vol. 24, no. 1, Art. no. 100658, 2026, doi: 10.1016/j.jgeb.2026.100658.

Reddy, C., Reddy, P. A., Janapati, H., Assiri, B., Shuaib, M., Alam, S., and Sheneamer, A., “A fine-tuned vision transformer-based enhanced multi-class brain tumor classification using MRI scan imagery,” Frontiers in Oncology, vol. 14, Art. no. 1400341, 2024, doi: 10.3389/fonc.2024.1400341.

Jiang, C., Wang, Y., Yuan, Q., et al., “A 3D medical image segmentation network based on gated attention blocks and dual-scale cross-attention mechanism,” Scientific Reports, vol. 15, Art. no. 6159, 2025, doi: 10.1038/s41598-025-90339-y.

Kaur, P., and Mahajan, P., “Detection of brain tumors using a transfer learning-based optimized ResNet152 model in MR images,” Computers in Biology and Medicine, vol. 188, Art. no. 109790, 2025, doi: 10.1016/j.compbiomed.2025.109790.

Hasan, M. J., Hasan, M., Akter, S., Mahi, A. B. S., and Uddin, M. P., “Enhancing brain tumor classification with a novel attention-based explainable deep learning framework,” Biomedical Signal Processing and Control, vol. 112, Art. no. 108636, 2026, doi: 10.1016/j.bspc.2025.108636.

Khan, S. U. R., Asif, S., Zhao, M., Zou, W., Li, Y., and Xiao, C., “ShallowMRI: A novel lightweight CNN with a novel attention mechanism for multi-brain-tumor classification in MRI images,” Biomedical Signal Processing and Control, vol. 111, Art. no. 108425, 2026, doi: 10.1016/j.bspc.2025.108425.

Uniyal, M., Saini, C., Singh, D. P., Ahmed, M. N., Hussain, M. R., Amarjeet, Sinha, A., and Srivastava, A., “Automated brain tumor detection using advanced deep learning models,” Discover Artificial Intelligence, vol. 6, Art. no. 1, 2026, doi: 10.1007/s44163-025-00753-4.

Batool, A., and Byun, Y.-C., “A lightweight multi-path convolutional neural network architecture using optimal feature selection for multiclass classification of brain tumors using magnetic resonance images,” Results in Engineering, vol. 25, Art. no. 104327, 2025, doi: 10.1016/j.rineng.2025.104327.

Anand, V., Gupta, S., Gupta, D., Gulzar, Y., Xin, Q., Juneja, S., Shah, A., and Shaikh, A., “Weighted average ensemble deep learning model for stratification of brain tumor in MRI images,” Diagnostics, vol. 13, no. 7, Art. no. 1320, 2023, doi: 10.3390/diagnostics13071320.

Zhang, Z., Sun, G., Zheng, K., Yang, J.-K., Zhu, X.-R., and Li, Y., “TC-Net: A joint learning framework based on CNN and vision transformer for multi-lesion medical image segmentation,” Computers in Biology and Medicine, vol. 161, Art. no. 106967, 2023, doi: 10.1016/j.compbiomed.2023.106967.

Zhao, J., and Li, S., “Evidence modeling for reliability learning and interpretable decision-making under multi-modality medical image segmentation,” Computerized Medical Imaging and Graphics, vol. 116, Art. no. 102422, 2024, doi: 10.1016/j.compmedimag.2024.102422.

L. Zhang, C. Lan, L. Fu, X. Mao, M. Zhang, “Segmentation of brain tumor MRI image based on improved attention module unet network”, Signal Image Video Process, Vol.17, no.5, 2023, doi.org/10.1007/s11760-022-02443-5

Zhou, M., Li, J., and Guo, Y., “Multi-level channel-spatial attention and light-weight scale-fusion network (MCSLF-Net): Multi-level channel-spatial attention and light-weight scale-fusion transformer for 3D brain tumor segmentation,” Quantitative Imaging in Medicine and Surgery, vol. 15, no. 7, pp. 6301–6325, 2025, doi: 10.21037/qims-2025-354.

Zhou, L., Jiang, Y., Li, W., Hu, J., and Zheng, S., “Shape-scale co-awareness network for 3D brain tumor segmentation,” IEEE Transactions on Medical Imaging, vol. 43, no. 7, pp. 2495–2508, 2024, doi: 10.1109/TMI.2024.3368531.

Hua, X., Du, Z., Ma, J., and Yu, H., “Multi-kernel cross-sparse graph attention convolutional neural network for brain magnetic resonance imaging super-resolution,” Biomedical Signal Processing and Control, vol. 96, Art. no. 106444, 2024, doi: 10.1016/j.bspc.2024.106444.

Wang, J., Lei, D., Zhang, Y., Yuan, J., Liu, C., Luo, B., Liu, Q., and Wang, G., “MedMamba: Multi-scale deformable attention via state-space models for robust medical image segmentation,” Biomedical Signal Processing and Control, vol. 112, Art. no. 108363, 2026, doi: 10.1016/j.bspc.2025.108363.

M. Jiang, F. Zhai, J. Kong, “A novel deep learning model DDU-net using edge features to enhance brain tumor segmentation on MR images”, Artificial Intelligence in Medicine, vol. 121, Art.no. 102180. ,2021, 10.1016/j.artmed.2021.102180.

H. Hedibi, M. Beladgham, A. Bouida, “A combined attention mechanism for brain tumor segmentation of lower-grade glioma in magnetic resonance images”, Computers in Biology and Medicine, vol.193, Art.no.110380, 2025, doi.org/10.1016/j.compbiomed.2025.110380.

Hasan, M. R., Rudra, S. M., Karmoker, N., Yousuf, M. A., Akhter, J., Al-Moisheer, A. S., Alyami, S. A., and Moni, M. A., “3D MRI reconstruction and brain tumor diagnosis using deep learning with explainable AI,” Expert Systems with Applications, vol. 315, Art. no. 131513, 2026, doi: 10.1016/j.eswa.2026.131513.

Li, C., Liu, Q., and Song, J., “Boundary-enhanced sparse transformer for generalizable and accurate medical image segmentation,” Scientific Reports, vol. 16, Art. no. 2191, 2026, doi: 10.1038/s41598-025-33686-0.