FracSegNet: A deep learning image segmentation model for medical images of eye diseases

Abstract

Retinal vessel segmentation is essential for the diagnosis and treatment planning of retinal diseases, yet remains challenging due to weak edges, tiny branches, and complex background textures. In this paper, we propose FracSegNet, a deep segmentation network that integrates fractional-order modeling at the preprocessing, feature extraction, and loss levels to improve the continuity and robustness of retinal vessel extraction. First, a Grünwald–Letnikov fractional differential operator is used to generate multi-directional edge responses, which are concatenated with the original fundus image to form an augmented multi-channel input. Second, adaptive fractional-order convolution blocks are embedded into a U-Net–like encoder–decoder architecture, where learnable order weights dynamically fuse integer-order and fractional-order responses, enabling simultaneous modeling of local details and long-range dependencies. Third, a composite loss is designed by combining Dice loss, total variation (TV) regularization, and a fractional gradient constraint that enforces consistency between the fractional-order gradients of the prediction and the ground truth. Experiments on the DRIVE, STARE, and CHASE DB1 datasets demonstrated that FracSegNet achieved competitive or superior performance compared with state-of-the-art methods, with F1 scores above 0.83 and clear improvements in edge continuity and fine-branch preservation. These results indicate that fractional-order modeling provides an effective and generalizable paradigm for segmenting weak edges and delicate vascular structures in medical images.

Keywords

Deep learning / Digital retinal images for vessel extraction / Fractional differential operator / Image segmentation / Medical image processing

References

[1] Wang Y, Wang J, Guo P. Eye-UNet. A UNet-based network with attention mechanism for low-quality human eye image segmentation. Signal Image Video Process. 2023; 17(4): 1097-1103. https://doi.org/10.1007/s11760-022-02316-x

[2] Chandra IS, Shastri RK, Kavitha D, Kumar KR, Manochitra S, Babu PB. CNN based color balancing and denoising technique for underwater images: CNN-CBDT. Meas Sens. 2023; 28: 100835. https://doi.org/10.1016/j.measen.2023.100835

[3] Ronneberger O, Fischer P, Brox T. U-Net: convolutional networks for biomedical image segmentation. In: MICCAI 2015. Cham: Springer; 2015: 234-241. https://doi.org/10.1007/978-3-319-24574-4_28

[4] Coelho C, Costa MFP, Ferrás LL. Fractional calculus meets neural networks for computer vision: a survey. AI. 2024; 5(3): 1391-1426. https://doi.org/10.3390/ai5030067

[5] Stember JN, Celik H, Krupinski E, Chang PD, Mutasa S, Lefkowicz A. Eye tracking for deep learning segmentation using convolutional neural networks. J Digit Imaging. 2019; 32(4): 597-604. https://doi.org/10.1007/s10278-019-00220-4

[6] Chen BZ, Liu YS, Zhang Z, et al. TransAttUnet: multi-level attention-guided U-Net with transformer for medical image segmentation. IEEE Trans Emerg Top Comput Intell. 2024; 8(1): 55-68. https://doi.org/10.1109/TETCI.2023.3309626

[7] Dosovitskiy A, Beyer L, Kolesnikov A, et al. An image is worth 16×16 words: transformers for image recognition at scale. In: Proc ICLR; 2021. Available at: https://openreview.net/forum?id=YicbFdNTTy

[8] Chen JN, Cheng J, Wang X, et al. TransUNet: transformers make strong encoders for medical image segmentation. arXiv. Preprint posted online February 8, 2021. Available at: https://arxiv.org/abs/2102.04306

[9] Azad R, Al-Antary MT, Heidari M, Merhof D. TransNorm: transformer provides a strong spatial normalization mechanism for a deep segmentation model. IEEE Access. 2022; 10: 108205-108215. https://doi.org/10.1109/ACCESS.2022.3211501

[10] Guo DF, Terzopoulos D. A transformer-based network for anisotropic 3D medical image segmentation. In: Proc ICPR. Milan, Italy; 2021: 8857-8861. https://doi.org/10.1109/ICPR48806.2021.9411990

[11] Zhang ZA, Wu CD, Coleman S, Tsakanikas V. DENSE-Inception U-net for medical image segmentation. Comput Methods Programs Biomed. 2020; 192: 105395. https://doi.org/10.1016/j.cmpb.2020.105395

[12] Liu HX, Fang JX, Zhang ZJ, et al. Localized edge-region-based active contour for medical image segmentation. IET Image Process. 2021; 15(7): 1567-1582. https://doi.org/10.1049/ipr2.12126

[13] Wang J, Wang W, Hu X, et al. Black-winged kite algorithm: a nature-inspired meta-heuristic for solving benchmark functions and engineering problems. Artif Intell Rev. 2024; 57(4): 1-53. https://doi.org/10.1007/s10462-024-10723-4

[14] Arora R, Kumar S, Singh P. Fractional-order image enhancement using deep learning for medical imaging applications. Neurocomputing. 2022; 501: 234-246

[15] Raubitzek S, Sabadas R, Latzko P. Combining fractional derivatives and convolutional neural networks for robust image denoising. Signal Process Image Commun. 2023; 28: 100835. https://doi.org/10.1016/j.imagecom.2022.116947

[16] Cao H, Wang Y, Chen J, et al. Swin-Unet: UNet-like pure transformer for medical image segmentation. In: Med Comput Vis Algorithms Big Data. Cham: Springer; 2022: 205-218. https://doi.org/10.1007/978-3-031-25066-8_9

[17] Chen B, Liu Y, Zhang Z, Lu G, Kong AWK. TransAttUnet: Multi-level attention-guided U-Net with transformer for medical image segmentation. IEEE Trans Emerg Top Comput Intell. 2023; 7(5): 1150-1163. https://doi.org/10.1109/TETCI.2023.3309626

[18] Jiang L, He Y, Zhang J. TransDualSegNet: Transformer dual-segmentation network for retinal vessel extraction in OCT images. Comput Biol Med. 2025; 173: 108023. https://doi.org/10.1016/j.compbiomed.2024.108023

[19] Dai J, Chen W, Chen R, et al. Research on task assignment algorithm of heterogeneous aircraft cooperative cluster in dynamic scene. Comput Electr Eng. 2023; 110: 108781. https://doi.org/10.1016/j.compeleceng.2023.108781

[20] Li X, Xie H, Zhang Z, Cai J. IterNet: Retinal image segmentation utilizing structural redundancy in vessel networks. In: Proc IEEE WACV; 2020. https://doi.org/10.1109/WACV45572.2020.9093621

[21] Zhou Z, Siddiquee MR, Tajbakhsh N, Liang J. UNet++: A nested U-Net architecture for medical image segmentation. In: Deep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support (DLMIA 2018); 2018: 3-11. https://doi.org/10.1007/978-3-030-00889-5_1

[22] Wu Y, Sang N, Xiang T. SCS-Net: scale and context sensitive network for retinal vessel segmentation. Med Image Anal. 2021; 70: 102025. https://doi.org/10.1016/j.media.2021.102025

[23] Mou L, Zhao Y, Chen L, et al. CS2-Net: Curvilinear structure segmentation network for retinal vessel extraction. Med Image Anal. 2021; 67: 101874. https://doi.org/10.1016/j.media.2020.101874

[24] Liu H, Fang J, Zhang Z. FR-UNet: Full-resolution network for medical image segmentation. IET Image Process. 2022; 16(7): 1567-1582. https://doi.org/10.1109/JBHI.2022.3188710

[25] Abdulsahib G, Al-Antary MT, Merhof D. Trainable filters for retinal vessel segmentation. Signal Process Image Commun. 2022; 114: 116947. https://doi.org/10.1016/j.imagecom.2022.116947

[26] Liu X, Wang Y, Zhang Z. WaveNet: Multi-frequency representation learning for retinal vessel segmentation. Comput Biol Med. 2023; 170: 106337. https://doi.org/10.1016/j.compbiomed.2022.106337

[27] Liu X, Fang J, Chen L. ResDO-UNet: Dual-orthogonal residual learning for retinal vessel segmentation. Biomed Signal Process Control. 2023; 84: 104810. https://doi.org/10.1016/j.bspc.2023.104810

[28] Li J, Gao G, Yang L, Bian G, Liu Y. DPF-Net: A dual-path progressive fusion network for retinal vessel segmentation. IEEE Trans Instrum Meas. 2023; 72: 1-17