The Emory Optic Nerve Head Atlas - Using 3D Anatomical Mapping to Study Optic Neuropathies with an Initial Focus on Glaucoma

TL;DR

This study introduces a novel 3D optic nerve head atlas created with AI to improve glaucoma diagnosis and severity assessment through consistent anatomical mapping and biomechanical analysis.

Contribution

The paper presents the first AI-driven 3D ONH atlas that enhances glaucoma diagnosis and severity classification by integrating structural and biomechanical features.

Findings

Atlas improves differentiation between healthy and mild glaucoma.

Atlas-adjusted RNFL thickness outperforms native measurements.

Strain-based features effectively classify glaucoma severity.

Abstract

Purpose: To develop the first 3D optic nerve head (ONH) atlas using AI-based registration and evaluate its use in: (1) atlas-adjusted retinal nerve fiber layer (RNFL) analysis for glaucoma diagnosis, and (2) strain-based assessment of glaucoma severity. Methods: Large-scale OCT datasets were registered using REFLECTIVITY-generated tissue segmentations. A healthy atlas (n=460) and glaucoma atlases for mild (n=852), moderate (n=640), and severe (n=546) disease were built using an AI-driven algorithm incorporating structural and biomechanical constraints. Atlas quality was evaluated using inter-layer contrast and agreement between warped scans and the template. Atlas-adjusted RNFL thickness was extracted by mapping fixed 1.5 mm BMO-based reference points from the atlas to each subject and compared with standard subject-specific RNFL measurements. Effective strain relative to each atlas was computed by morphing subjects to the atlas template. Strain and thickness features were used to classify glaucoma severity. Results: The healthy atlas captured normative ONH anatomy, while glaucoma atlases showed stage-dependent changes. Registration accuracy was high (NCC=0.86+-0.05; Dice=0.90+-0.02), with strong inter-layer contrast (0.40+-0.03). Atlas-adjusted RNFL thickness improved separation between healthy and mild glaucoma (Cohen's d=0.77 vs. 0.72). Classification using atlas-adjusted profiles outperformed native measurements (AUC=0.771 vs. 0.757). Mean effective strain differed significantly across severity groups (ANOVA p<0.05), and CNNs trained on strain maps achieved strong classification (AUC=0.79). Conclusion: The ONH atlas provides a population-derived structural reference enabling anatomically consistent thickness mapping and strain-based characterization. It enhances diagnostic discrimination, supports severity classification, and offers a framework to study glaucomatous biomechanics.