Retinal Nerve Fiber Layer Defect Detection with Inexact Position-Wise Sector Contrastive Learning on Retinal Fundus Images
Source abstract
Abstract
Retinal nerve fiber layer defect (RNFLD) is a critical early indicator of glaucoma. Existing RNFLD detection methods underutilize clinical anatomical priors, struggle with subtle or diffuse lesions, and rely heavily on high-quality annotations. To address these issues, this paper proposes Inexact Position-Wise Sector Contrastive Learning (IPSCL) for RNFLD detection on retinal fundus images. IPSCL combines position guidance, inexact supervised learning based on a novel Radial Cup–Disc Ratio (RCDR), and sector-level contrastive learning to leverage anatomical correlations and enhance fine-grained feature discrimination. Experiments on the original RNFLD dataset and its extended version show that IPSCL outperforms state-of-the-art methods, achieving sector-level F1 score of 83.84% and image-level F1 score of 94.98%. The method effectively detects subtle and diffuse defects while reducing annotation dependence. This work provides a reliable and interpretable solution for clinical RNFLD detection.
Source: crossref · Quality: baseline
Research context
OpenAlex topics: Retinal Imaging and Analysis · Retinal Diseases and Treatments · Retinal and Macular Surgery
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
Affiliations
School of Intelligent Science and Technology, Beijing University of Civil Engineering and Architecture, Beijing 102616, China
References (41)
Show source-provided references
1. Global estimates of visual impairment: 201010.1136/bjophthalmol-2011-300539 ↗
2. The pathophysiology and treatment of glaucoma: A review10.1001/jama.2014.3192 ↗
3. Quantification of retinal nerve fibre layer thickness on optical coherence tomography with a deep learning segmentation-free approach10.1038/s41598-019-57196-y ↗
4. Automated segmentation by pixel classification of retinal layers in ophthalmic OCT images10.1364/BOE.2.001743 ↗
5. Learning layer-specific edges for segmenting retinal layers with large deformations10.1364/BOE.7.002888 ↗
6. Retinal layers segmentation using fully convolutional network in OCT images
7. Global, regional and national burden of Glaucoma: An update analysis from the Global Burden of Disease Study 201910.1007/s10792-024-03222-6 ↗
8. Optical coherence tomography versus optic disc photo assessment in glaucoma screening10.1097/IJG.0000000000002392 ↗
9. Automated retinal nerve fiber layer defect detection using fundus imaging in glaucoma10.1016/j.compmedimag.2018.02.006 ↗
10. Early detection of glaucoma through retinal nerve fiber layer analysis using fractal dimension and texture feature10.15623/ijret.2014.0310023 ↗
11. Detection of retinal nerve fiber layer defects on retinal fundus images for early diagnosis of glaucoma10.1117/1.3322388 ↗
12. Deep convolutional neural network-based patch classification for retinal nerve fiber layer defect detection in early glaucoma10.1117/1.JMI.5.4.044003 ↗
13. Panda, R., Puhan, N.B., Rao, A., Padhy, D., and Panda, G. (2017). Recurrent neural network based retinal nerve fiber layer defect detection in early glaucoma. 2017 IEEE 14th International Symposium on Biomedical Imaging (ISBI 2017), IEEE.10.1109/ISBI.2017.7950614 ↗
14. Ding, F., Yang, G., Ding, D., and Cheng, G. (2020). Retinal nerve fiber layer defect detection with position guidance. International Conference on Medical Image Computing and Computer-Assisted Intervention, Springer.10.1007/978-3-030-59722-1_72 ↗
15. Na, K.I., Jeoung, J.W., Lee, W.J., Kim, Y.K., Kim, C.Y., Park, K.H., and Epidemiologic Survey Committee of the Korean Ophthalmologic Society (2017). Prevalence of retinal nerve fiber layer defects: The Korea National Health and Nutrition Examination Survey 2008–2012. PLoS ONE, 12.10.1371/journal.pone.0186032 ↗
16. Association of progressive optic disc tilt with development of retinal nerve fibre layer defect in children with large cup-to-disc ratio10.1136/bjophthalmol-2021-320029 ↗
17. Advanced AI strategies for accurate detection and localization of retinal nerve fiber layer defects: Employing multifaceted analysis across various color spaces10.1007/s13721-025-00507-3 ↗
18. A hierarchical deep learning approach with transparency and interpretability based on small samples for glaucoma diagnosis10.1038/s41746-021-00417-4 ↗
19. Diffmic-v2: Medical image classification via improved diffusion network10.1109/TMI.2025.3530399 ↗
20. Medical image classification with KAN-integrated transformers and dilated neighborhood attention10.1016/j.asoc.2025.114045 ↗
21. Ranking of optic disc variables for detection of glaucomatous optic nerve damage
22. Superpixel classification based optic disc and optic cup segmentation for glaucoma screening10.1109/TMI.2013.2247770 ↗
23. Noor, N.M., Khalid, N.E.A., and Ariff, N.M. (2013). Optic cup and disc color channel multi-thresholding segmentation. 2013 IEEE International Conference on Control System, Computing and Engineering, IEEE.10.1109/ICCSCE.2013.6720022 ↗
24. Ayub, J., Ahmad, J., Muhammad, J., Aziz, L., Ayub, S., Akram, U., and Basit, I. (2016). Glaucoma detection through optic disc and cup segmentation using K-mean clustering. 2016 International Conference on Computing, Electronic and Electrical Engineering (ICE Cube), IEEE.10.1109/ICECUBE.2016.7495212 ↗
25. Joint optic disc and cup segmentation based on multi-label deep network and polar transformation10.1109/TMI.2018.2791488 ↗
26. Ronneberger, O., Fischer, P., and Brox, T. (2015). U-net: Convolutional networks for biomedical image segmentation. Medical Image Computing and Computer-Assisted Intervention—MICCAI 2015: 18th International Conference, Munich, Germany, 5–9 October 2015, Springer. Proceedings, Part III 18.
27. JointRCNN: A region-based convolutional neural network for optic disc and cup segmentation10.1109/TBME.2019.2913211 ↗
28. Yin, P., Wu, Q., Xu, Y., Min, H., Yang, M., Zhang, Y., and Tan, M. (2019). PM-Net: Pyramid multi-label network for joint optic disc and cup segmentation. International Conference on Medical Image Computing and Computer-Assisted Intervention, Springer.10.1007/978-3-030-32239-7_15 ↗
29. Ding, F., Yang, G., Liu, J., Wu, J., Ding, D., Xv, J., Cheng, G., and Li, X. (2019). Hierarchical attention networks for medical image segmentation. arXiv.10.1007/978-3-030-59710-8_25 ↗
30. Metadata-enhanced contrastive learning from retinal optical coherence tomography images10.1016/j.media.2024.103296 ↗
31. Learning better contrastive view from radiologist’s gaze10.1016/j.patcog.2025.111350 ↗
32. Spatial-aware contrastive learning for cross-domain medical image registration10.1002/mp.17311 ↗
33. Multi-modal contrastive mutual learning and pseudo-label re-learning for semi-supervised medical image segmentation10.1016/j.media.2022.102656 ↗
34. PatchCL-AE: Anomaly detection for medical images using patch-wise contrastive learning-based auto-encoder10.1016/j.compmedimag.2024.102366 ↗
35. Yang, Y., Zhuang, J., Sun, G., Wang, R., and Su, J. (2025). Boundary-guided Contrastive Learning for Semi-supervised Medical Image Segmentation. IEEE Transactions on Medical Imaging, IEEE.10.1109/TMI.2025.3556482 ↗
36. A brief introduction to weakly supervised learning10.1093/nsr/nwx106 ↗
37. Anklin, V., Pati, P., Jaume, G., Bozorgtabar, B., Foncubierta-Rodriguez, A., Thiran, J.P., Sibony, M., Gabrani, M., and Goksel, O. (2021). Learning whole-slide segmentation from inexact and incomplete labels using tissue graphs. Medical Image Computing and Computer Assisted Intervention—MICCAI 2021: 24th International Conference, Strasbourg, France, 27 September–1 October 2021, Springer. Proceedings, Part II 24.10.1007/978-3-030-87196-3_59 ↗
38. Short-wavelength automated perimetry results are correlated with optical coherence tomography retinal nerve fiber layer thickness measurements in glaucomatous eyes10.1016/j.ophtha.2004.04.017 ↗
39. OCT structural abnormality detection in glaucoma using topographically correspondent rim and retinal nerve fiber layer criteria10.1016/j.ajo.2019.12.020 ↗
40. Dimensionality reduction by learning an invariant mapping10.1109/CVPR.2006.100 ↗
41. The ISNT rule and differentiation of normal from glaucomatous eyes10.1001/archopht.124.11.1579 ↗