Clinical Report: RADAR Framework for 3D Retinal Microvasculature Mapping in Diabetic Retinopathy

Overview

RADAR is an unsupervised computational framework enabling 3D segmentation and quantification of retinal microvasculature from OCT angiography data without manual annotation. Validated in healthy and early diabetic retinopathy patients, RADAR outperforms standard tools and reveals layer-specific vascular changes and volumetric biomarkers.

Background

Three-dimensional mapping of retinal microvasculature is critical for monitoring systemic vascular health and early detection of diseases such as diabetic retinopathy. Traditional methods depend on manual annotation or deep learning models that require extensive training and may lack generalizability. Optical coherence tomography angiography (OCTA) provides detailed vascular imaging, but conventional 2D projections obscure complex morphological alterations. There is a need for scalable, annotation-free tools to accurately reconstruct and analyze retinal vascular networks in 3D.

Data Highlights

The RADAR framework integrates adaptive physics-aware denoising with topology-preserving centerline extraction to reconstruct complex 3D vascular networks. Validation studies demonstrated superior segmentation performance compared to standard methods. Quantitative analysis identified increased vessel tortuosity and compensatory remodeling patterns in diabetic eyes, with precise extraction of volumetric biomarkers such as vessel length and branching complexity. Layer-specific morphological alterations were resolved that are not visible in conventional 2D OCTA projections.

Key Findings

• RADAR enables annotation-free 3D segmentation of retinal microvasculature from OCTA data.
• The method outperforms standard segmentation tools in accuracy and detail.
• It reveals layer-specific vascular morphological changes obscured in 2D projections.
• Quantitative biomarkers extracted include vessel length, branching complexity, and tortuosity.
• Distinct compensatory remodeling patterns were identified in early diabetic retinopathy patients.
• The framework supports scalable, longitudinal assessment of ocular and systemic vascular diseases.

Clinical Implications

RADAR provides clinicians with a precise, scalable tool for early detection and monitoring of microvascular changes in diabetic retinopathy and other systemic vascular conditions. Its annotation-free approach reduces reliance on manual labeling and enhances reproducibility. The ability to extract volumetric biomarkers and resolve layer-specific alterations may improve disease staging and guide therapeutic interventions.

Conclusion

The RADAR framework represents a significant advancement in 3D retinal microvasculature imaging, enabling detailed, annotation-free analysis that can enhance early diagnosis and longitudinal monitoring of diabetic retinopathy and related vascular diseases.

References

1. Liew et al. 2008 -- Retinal vascular imaging: a new tool in microvascular disease research
2. Cheung et al. 2012 -- Retinal microvasculature as a model to study the manifestations of hypertension
3. London et al. 2013 -- The retina as a window to the brain—from eye research to CNS disorders
4. Zhang et al. 2021 -- Early detection of microvascular impairments with optical coherence tomography angiography in diabetic patients without clinical retinopathy: a meta-analysis
5. Isaac et al. 2025 -- Effect of disease duration on foveal microvasculature assessed by OCTA in type 2 diabetes mellitus without clinical diabetic retinopathy
6. Zhang et al. 2025 -- Recent advances and applications of optical coherence tomography angiography in diabetic retinopathy