12, December 2025

HYBRID SWINUNETR-MLP FRAMEWORK FOR AUTOMATED LEUKEMIA DETECTION IN MICROSCOPIC BLOOD CELLS

VOLUME-11; ISSUE-12; DEC-2025

Author(s): Kumbha Praveen Kumar¹, S. Swarnalatha²

Authors Affiliations:

1.PG Student, Department of Electronic Communication and Engineering, Sri Venkateshwara University College of Engineering, Tirupati, Andhra Pradesh, India

2.Professor, Department of Electronic Communication and Engineering, Sri Venkateshwara University College of Engineering, Tirupati, Andhra Pradesh, India

DOIs:10.2015/IJIRMF/202512002     |     Paper ID: IJIRMF202512002


Abstract
Keywords
Cite this Article/Paper as
References

The automatic recognition of leukaemia on microscopic blood smear images has become a significant point in the application of deep learning to medical diagnostics. Traditional convolutional neural network (CNN) models have demonstrated good performance in the classification of blood cell abnormalities, yet tend to have poor contextualization and localisation of malignant areas. The paper presents a new Hybrid SwinUNETR-MLP, at the same time combining a transformer-based encoder-decoder segmentation network with a lightweight multilayer perceptron (MLP) classifier to simultaneously segment cells and classify diseases by their stage. The SwinUNETR backbone is an effective hierarchical spatial dependency amount of the shifted-window self-attention, whereas the appended MLP head is able to do the global reasoning using the features of segmentation to generate leukaemia subtypes, such as Pro, Pre, Early, and Benign cells. This system is developed in Python with PyTorch and MONAI libraries with a graphical user interface (GUI) to aid in real-time diagnosis. It shows better results in segmentation and classification accuracy as experimental validation on a publicly available microscopic blood-smear dataset achieves a Dice similarity coefficient of 0.93 and an overall classification accuracy of 97.2 per cent, being among the best U-Net and ResNet-based models. The framework provides end-to-end automation, including image selection to diagnostic visualisation, and makes it interpretable and clinically applicable. The hybrid architecture of the proposed model increases the generalisation and robustness, which proves the possibility of the model as a potentially effective computer-aided diagnostic tool in the biomedical imaging and electronic communication systems with an integrated approach to intelligent healthcare infrastructure.

Leukaemia Detection, SwinUNETR Multilayer Perceptron (MLP) Transformer Medical Image Segmentation, Deep Learning, Hybrid Architecture, Biomedical Signal Processing.

Kumbha Praveen Kumar¹, S. Swarnalatha² (2025); HYBRID SWINUNETR-MLP FRAMEWORK FOR AUTOMATED LEUKEMIA DETECTION IN MICROSCOPIC BLOOD CELLS, International Journal for Innovative Research in Multidisciplinary Field, ISSN(O): 2455-0620, Vol-11, Issue-12, Pp.9-19          Available on –   https://www.ijirmf.com/

  1. Ronneberger, O., Fischer, P., & Brox, T. (2015). U-Net: Convolutional networks for biomedical image segmentation. MICCAI, 234–241.
  2. Vaswani, A. et al. (2017). Attention is all you need. Advances in Neural Information Processing Systems (NeurIPS).
  3. Dosovitskiy, A. et al. (2021). An Image is Worth 16×16 Words: Transformers for Image Recognition at Scale. ICLR.
  4. Hatamizadeh, A., Nath, V., Tang, Y., et al. (2022). Swin UNETR: Swin Transformers for semantic segmentation of brain tumours. MICCAI Workshop on BrainLes.
  5. He, K., Zhang, X., Ren, S., & Sun, J. (2016). Deep residual learning for image recognition. CVPR, 770–778.
  6. Huang, G., Liu, Z., Van Der Maaten, L., & Weinberger, K. Q. (2017). Densely connected convolutional networks. CVPR.
  7. Jha, D., Smedsrud, P. H., et al. (2020). ResUNet++: An advanced architecture for medical image segmentation. IEEE Access, 8, 41758–41769.
  8. Ciga, O., Xu, T., & Martel, A. L. (2021). Self-supervised contrastive learning for digital histopathology. Medical Image Analysis, 71, 102058.
  9. Mahmood, F., Borders, D., Chen, R., McKay, G. N., & Durr, N. J. (2020). Deep learning-based segmentation and classification of leukocytes. IEEE Transactions on Biomedical Engineering, 67(8), 2354–2368.
  10. Rehman, A., et al. (2018). Classification of acute lymphoblastic leukaemia using deep learning. Microscopy Research and Technique, 81(11), 1310–1317.
  11. Razzak, M. I., Imran, M., & Xu, G. (2019). Efficient CNN-based blood cell classification. IEEE Access, 7, 68375–68382.
  12. Chen, J., Lu, Y., Yu, Q., et al. (2021). TransUNet: Transformers make strong encoders for medical image segmentation. arXiv:2102.04306.
  13. Wang, J., Zhang, T., Zhang, H., et al. (2023). Multi-task learning in medical imaging: A review. Medical Image Analysis, 87, 102797.
  14. Nwankpa, C., Ijomah, W., Gachagan, A., & Marshall, S. (2018). Activation functions: Comparison and trends in deep learning. arXiv preprint arXiv:1811.03378.
  15. Zhang, X., Han, J., & Zhao, Y. (2022). Edge intelligence for healthcare monitoring systems. IEEE Internet of Things Journal, 9(13), 11254–11267.
Post Views: 76

Download Full Paper

Download PDF No. of Downloads:13 | No. of Views: 68