Citation :

Sudhakar Jyothula, Ch Sekhar, I Lakshmi Manikyamba, K Nagaraju, Hari Jyothula, Subbarao P, "HALS-Net: Hybrid Adaptive Level Set Network for Precision Biomedical Image Segmentation with Neural Attention-Guided Contour Evolution," International Journal of Electronics and Communication Engineering, vol. 13, no. 4, pp. 277-284, 2026. Crossref, https://doi.org/10.14445/23488549/IJECE-V13I4P123

Abstract

Biomedical image segmentation has long been a challenging issue in computational pathology, radiology, and ophthalmology due to the complicated tissue morphologies, intensity inhomogeneity, and weak boundary gradients that usually make traditional methods fail. Level Set Methods (LSMs) have been a long-standing mathematically elegant representation for segmentation of images because LSMs are inherently able to continuously represent shape topology, but conventional formulations are susceptible to initialization and slow convergence, as well as limited in their capability to exploit high-level semantic information. This paper proposes a new paradigm called HALS-Net (Hybrid Adaptive Level Set Network), which seamlessly incorporates the wisdom of classical layer-based level set theory and that of attention mechanisms in deep neural networks, attaining state-of-the-art for biomedical segmentation in images. The proposed model establishes a spatially varying energy functional in which region-based and edge-based driving forces are dynamically balanced by means of an attention weight controller, for the purpose of robust contour evolution under severe intensity inhomogeneity and noise.
Architecture of HALS-Net: The model is organized by four coupled modules, which are (i) an anisotropic diffusion-based preprocessing architecture subject to Contrast-Limited Adaptive Histogram Equalization (CLAHE), (ii) a multi-scale deep feature extraction backbone upon an attention U-Net encoder, (iii) a hybrid adaptive level set evolution module that combines Chan-Vese global region fitting and Local Binary Fitting (LBF) energy, and geodesic active contour edge attraction through neural attention-weighted coefficients, and (iv) curvature-based regularization with distance-preserving re-initialization. Comprehensive experiments on five benchmark datasets (i.e., Brats 2024, DRIVE, ISIC 2024, MoNuSeg, and Montgomery County Chest X-ray) are conducted to demonstrate that HALS-Net outperforms present state-of-the-art deep learning-based models and level set methods by improving the mean Dice Similarity Coefficient (DSC) ranging from 3.2% to 7.8%, while achieving the computational efficiency required for clinical application. The proposed methodology paves the way to a new paradigm in physics-informed deep segmentation by integrating differentiable Partial Differential Equation (PDE) solvers into end-to-end trainable neural architectures.

Keywords

Level Set Methods, Biomedical Image Segmentation, Attention Mechanisms, Variational Energy Functionals, Deep Learning, Contour Evolution.

References

1. Stanley Osher, and James A Sethian, “Fronts Propagating with Curvature-Dependent Speed: Algorithms based on Hamilton-Jacobi Formulations,” Journal of Computational Physics, vol. 79, no. 1, pp. 12-49, 1988.
   [CrossRef] [Google Scholar] [Publisher Link]
2. Vicent Caselles, Ron Kimmel, and Guillermo Sapiro, “Geodesic Active Contours,” International Journal of Computer Vision, vol. 22, pp. 61-79, 1997.
   [CrossRef] [Google Scholar] [Publisher Link]
3. T.F. Chan, and L.A. Vese, “Active Contours without Edges,” IEEE Transactions on Image Processing, vol. 10, no. 2, pp. 266-277, 2001.
   [CrossRef] [Google Scholar] [Publisher Link]
4. Chunming Li et al., “Minimization of Region-Scalable Fitting Energy for Image Segmentation,” IEEE Transactions on Image Processing, vol. 17, no. 10, pp. 1940-1949, 2008.
   [CrossRef] [Google Scholar] [Publisher Link]
5. Chunming Li et al., “Distance Regularized Level Set Evolution and Its Application to Image Segmentation,” IEEE Transactions on Image Processing, vol. 19, no. 12, pp. 3243-3254, 2010.
   [CrossRef] [Google Scholar] [Publisher Link]
6. Olaf Ronneberger, Philipp Fischer, and Thomas Broxy, “U-Net: Convolutional Networks for Biomedical Image Segmentation,” 18^th International Conference Proceedings Medical Image Computing and Computer-Assisted Intervention – MICCAI 2015, Munich, Germany, pp. 234-241, 2015. [CrossRef] [Google Scholar] [Publisher Link]
7. Ozan Oktay et al., “Attention U-Net: Learning Where to Look for the Pancreas,” arXiv preprint, pp. 1-10, 2018.
   [CrossRef] [Google Scholar] [Publisher Link]
8. Zongwei Zhou et al., “UNet++: Redesigning Skip Connections to Exploit Multiscale Features in Image Segmentation,” IEEE Transactions on Medical Imaging, vol. 39, no. 6, pp. 1856-1867, 2020.
   [CrossRef] [Google Scholar] [Publisher Link]
9. Jieneng Chen et al., “TransUNet: Transformers Make Strong Encoders for Medical Image Segmentation,” arXiv preprint, pp. 1-13, 2021.
   [CrossRef] [Google Scholar] [Publisher Link]
10. Fabian Isensee et al., “nnU-Net: A Self-Configuring Method for Deep Learning-based Biomedical Image Segmentation,” Nature Methods, vol. 18, pp. 203-211, 2021.
    [CrossRef] [Google Scholar] [Publisher Link]
11. Ali Hatamizadeh et al., “Swin UNETR: Swin Transformers for Semantic Segmentation of Brain Tumors in MRI Images,” 7^th International Conference Proceedings Brainlesion: Glioma, Multiple Sclerosis, Stroke and Traumatic Brain Injuries, pp. 272-284, 2022.
    [CrossRef] [Google Scholar] [Publisher Link]
12. Simon Graham et al., “Hover-Net: Simultaneous Segmentation and Classification of Nuclei in Multi-tissue Histology Images,” Medical Image Analysis, vol. 58, pp. 1-18, 2019.
    [CrossRef] [Google Scholar] [Publisher Link]
13. Junde Wu et al., “MedSegDiff: Medical Image Segmentation with Diffusion Probabilistic Model,” Proceedings of Medical Imaging with Deep Learning (MIDL), vol. 227, pp. 1623-1639, 2024.
    [Google Scholar] [Publisher Link]
14. Ali Hatamizadeh et al., “Deep Active Lesion Segmentation,” Conference Proceedings 10^th International Machine Learning in Medical Imaging, Shenzhen, China, pp. 98-105, 2019.
    [CrossRef] [Google Scholar] [Publisher Link]
15. Junde Wu et al., “MedSegDiff-V2: Diffusion-Based Medical Image Segmentation with Transformer,” Proceedings of the 38^th AAAI Conference on Artificial Intelligence, vol. 38, no. 6, pp. 6030-6038, 2024.
    [CrossRef] [Google Scholar] [Publisher Link]
16. Mo Zhang, Bin Dong, and Quanzheng Li, “Deep Active Contour Network for Medical Image Segmentation,” 23^rd International Conference Proceedings Medical Image Computing and Computer Assisted Intervention (MICCAI), Lima, Peru, pp. 321-331, 2020.
    [CrossRef] [Google Scholar] [Publisher Link]
17. Junlong Cheng et al., “SAM-Med2D,” arXiv preprint, pp. 1-16, 2023.
    [CrossRef] [Google Scholar] [Publisher Link]
18. Wentao Liu et al., “Full-Resolution Network and Dual-Threshold Iteration for Retinal Vessel and Coronary Angiograph Segmentation,” IEEE Journal of Biomedical and Health Informatics, vol. 26, no. 9, pp. 4623-4634, 2022.
    [CrossRef] [Google Scholar] [Publisher Link]
19. Fabian Hörst et al., “CellViT: Vision Transformers for Precise Cell Segmentation and Classification,” Medical Image Analysis, vol. 94, pp. 1-16, 2024.
    [CrossRef] [Google Scholar] [Publisher Link]
20. Noel Codella et al., “Skin Lesion Analysis toward Melanoma Detection 2018: A Challenge Hosted by the International Skin Imaging Collaboration (ISIC),” arXiv preprint, pp. 1-12, 2019.
    [CrossRef] [Google Scholar] [Publisher Link]
21. Bjoern Menze et al., “The 2024 Brain Tumor Segmentation (BraTS) Challenge: Glioma Segmentation on Post-treatment MRI,” arXiv preprint, pp. 1-10, 2024.
    [CrossRef] [Google Scholar] [Publisher Link]
22. Xiaoling Hu et al., “Topology-Preserving Deep Image Segmentation,” 33^rd Conference on Advances in Neural Information Processing Systems (NeurIPS), Vancouver, Canada, vol. 32, pp. 1-12, 2019.
    [Google Scholar] [Publisher Link]
23. J. Staal et al., “Ridge-based Vessel Segmentation in Color Images of the Retina,” IEEE Transactions on Medical Imaging, vol. 23, no. 4, pp. 501-509, 2004.
    [CrossRef] [Google Scholar] [Publisher Link]