DeepLabV3+ with ResNeXt50 Backbone for Automated Brain MRI Tumor Segmentation

Section: Research Article
Issue
Vol. 35 No. 3 (2026): Volume 35, Issue 3
Pages
15-28

Keywords:

Brain Tumor Segmentation, Deep Learning, Medical Imaging, Desktop Application, Clinical Decision Support, MRI Analysis

Abstract

This essay describes the design and testing of a desktop computer-based automated brain tumour segmentation system for MRI images, which will support the broad adoption of advanced deep learning models in clinical practice. The proposed system integrates a DeepLabV3+ framework with a ResNeXt50 backbone and a U-Net complementary structure, and is trained on a hybrid loss function that combines Dice loss and binary cross-entropy. It was trained and evaluated using benchmark brain MRI data, such as the BraTS 2020 dataset, with a standardised preprocessing and augmentation pipeline. In contrast to most research-only implementations, the proposed work focuses on practical deployment by incorporating the entire workflow, i.e., data preprocessing, model training, inference, and quantitative evaluation, in an integrated desktop environment suitable for clinical settings. The system will facilitate effective segmentation while maintaining high diagnostic reliability. The quantitative analysis yields promising results, with a mean Dice coefficient of 0.89 ± 0.04, a sensitivity of 81.5%, a specificity of 95.9%, and an AUC-ROC greater than 0.90. It has a mean inference time of 0.65 seconds per image and moderate memory requirements, making it easy to integrate into a typical hospital computing infrastructure. The findings suggest that the tool offers a good compromise between segmentation accuracy, computational efficiency, and clinical usability. This balance may support feasible AI-based decision support in neuro-oncology.

References

  1. A. B. Abdusalomov, M. Mukhiddinov, and T. K. Whangbo, “Brain Tumor Detection Based on Deep Learning Approaches and Magnetic Resonance Imaging.,” Cancers (Basel)., vol. 15, no. 16, Aug. 2023.
  2. N. Musthafa, Q. A. Memon, and M. M. Masud, “Advancing Brain Tumor Analysis: Current Trends, Key Challenges, and Perspectives in Deep Learning-Based Brain MRI Tumor Diagnosis,” Eng, vol. 6, no. 5. p. 82, 2025.
  3. A. Adadi, “A survey on data‐efficient algorithms in big data era,” J. Big Data, vol. 8, Jan. 2021.
  4. C. Sarkar et al., “Artificial Intelligence and Machine Learning Technology Driven Modern Drug Discovery and Development,” International Journal of Molecular Sciences, vol. 24, no. 3. p. 2026, 2023.
  5. M. A. Hamed et al., “Prostate cancer-derived extracellular vesicles metabolic biomarkers: Emerging roles for diagnosis and prognosis,” J. Control. Release, vol. 371, pp. 126–145, 2024.
  6. A. Roy Choudhury, R. Vanguri, S. R. Jambawalikar, and P. Kumar, “Segmentation of Brain Tumors Using DeepLabv3+,” in Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries: 4th International Workshop, BrainLes 2018, Held in Conjunction with MICCAI 2018, Granada, Spain, September 16, 2018, Revised Selected Papers, Part II, 2018, pp. 154–167.
  7. F. Isensee, F. J. Paul, P. M. Full, and P. Vollmuth, “nnU-Net for Brain Tumor Segmentation,” arXiv:2011.00848v1, pp. 1–15, 2020.
  8. S. Ahuja, B. Panigrahi, and T. Gandhi, Fully automatic brain tumor segmentation using DeepLabv3+ with variable loss functions. 2021.
  9. T. Saeed et al., “Neuro-XAI: Explainable deep learning framework based on deeplabV3+ and bayesian optimization for segmentation and classification of brain tumor in MRI scans,” J. Neurosci. Methods, vol. 410, p. 110247, 2024.
  10. M. Abrar, A. Salam, F. Ullah, F. Ullah, and A. S. Al Ghamdi, “Enhancing brain tumor segmentation using attention based convolutional UNet on MRI images.,” Sci. Rep., vol. 15, no. 1, p. 36603, Oct. 2025.
  11. M. A. Talukder, M. A. Layek, M. A. Hossain, M. A. Islam, M. Nur-e-Alam, and M. Kazi, “ACU-Net: Attention-based convolutional U-Net model for segmenting brain tumors in fMRI images,” Digit. Heal., vol. 11, p. 20552076251320290, 2025.
  12. M. Xiong, A. Wu, Y. Yang, and Q. Fu, “Efficient Brain Tumor Segmentation for MRI Images Using YOLO-BT,” Sensors, vol. 25, no. 12, 2025.
  13. Y. Fan et al., “RMAP-ResNet: Segmentation of brain tumor OCT images using residual multicore attention pooling networks for intelligent minimally invasive theranostics,” Biomed. Signal Process. Control, vol. 90, p. 105805, 2024.
  14. A. F. GARRIDO, “A COMPARISON STUDY OF U-NET BASED METHODS FOR BRAIN TUMOR SEGMENTATION,” 2023.
  15. M. Obayya et al., “A novel U-net model for brain tumor segmentation from MRI images,” Alexandria Eng. J., vol. 126, pp. 220–230, 2025.
  16. A. Pourmahboubi, N. Arsalani Saeed, and H. Tabrizchi, “A brain tumor segmentation enhancement in MRI images using U-Net and transfer learning,” BMC Med. Imaging, vol. 25, no. 1, p. 307, 2025.
  17. J. Zhu, R. Zhang, and H. Zhang, “An MRI brain tumor segmentation method based on improved U-Net,” vol. 21, no. December, pp. 778–791, 2023.
  18. S. Appavu and J. S. Bobby, “DSRA-UNet-STB : A DUAL-SCALE RESIDUAL ATTENTION U-NET WITH SWIN TRANSFORMER BLOCK FOR BRAIN TUMOR SEGMENTATION,” vol. 11, no. 1, pp. 117–128, 2025.

Author

Harith Safwan Izzaldeen

Department of Graduate Students, Faculty of Engineering, Islamic University of Lebanon

Identifiers

DOI: 10.33899/jes.v35i3.60403

Download this PDF file
PDF

Statistics

How to Cite

[1]
“DeepLabV3+ with ResNeXt50 Backbone for Automated Brain MRI Tumor Segmentation”, JES, vol. 35, no. 3, pp. 15–28, Jul. 2026, doi: 10.33899/jes.v35i3.60403.
Copyright and Licensing
Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 International License.

How to Cite

[1]
“DeepLabV3+ with ResNeXt50 Backbone for Automated Brain MRI Tumor Segmentation”, JES, vol. 35, no. 3, pp. 15–28, Jul. 2026, doi: 10.33899/jes.v35i3.60403.