Physics-based generation of multilayer corneal OCT data via Gaussian modeling and MCML for AI-driven diagnostic and surgical guidance applications

TL;DR

The paper tackles the data scarcity problem in corneal OCT by introducing a configurable physics-based synthetic pipeline that generates high-resolution B-scans with pixel-perfect five-layer masks and aligned optical maps. It combines a Gaussian-surface multilayer corneal geometry model with MCML forward light transport, augmented by confocal PSF gating and sensitivity roll-off, to produce over $10,000$ samples at $1024×1024$ resolution. Geometry is parameterized to span healthy and keratoconus-like morphologies, using $y_{base}(x)$ and deformation Δy(x)$ to model curvature variations. The synthetic dataset enables controlled benchmarking for AI, demonstrated via diffusion-based joint optical-map/structural OCT reconstruction and three-class segmentation, with improvements over baselines and near-ceiling performance on labelled layers. This resource provides a scalable, ground-truth foundation for robust AI-driven diagnostic and surgical guidance in image-guided ophthalmology.

Abstract

Training deep learning models for corneal optical coherence tomography (OCT) imaging is limited by the availability of large, well-annotated datasets. We present a configurable Monte Carlo simulation framework that generates synthetic corneal B-scan optical OCT images with pixel-level five-layer segmentation labels derived directly from the simulation geometry. A five-layer corneal model with Gaussian surfaces captures curvature and thickness variability in healthy and keratoconic eyes. Each layer is assigned optical properties from the literature and light transport is simulated using Monte Carlo modeling of light transport in multi-layered tissues (MCML), while incorporating system features such as the confocal PSF and sensitivity roll-off. This approach produces over 10,000 high-resolution (1024x1024) image-label pairs and supports customization of geometry, photon count, noise, and system parameters. The resulting dataset enables systematic training, validation, and benchmarking of AI models under controlled, ground-truth conditions, providing a reproducible and scalable resource to support the development of diagnostic and surgical guidance applications in image-guided ophthalmology.

Figures (7)

• Figure 1: (a) Semi-circular baseline curvature used as the first-order corneal support profile (red dashed curve). (b) Clinical corneal OCT B-scan with the fitted semi-circular baseline overlaid (red dashed curve) to illustrate dome-shape consistency. Panel (b) is from the public dataset https://doi.org/10.6084/m9.figshare.25952845sun2024oct and was cropped and annotated for visualization.
• Figure 2: Representative simulated corneal OCT B-scans generated by our model: a healthy cornea (left) and a keratoconic cornea (right), where a localized anterior steepening is introduced to mimic keratoconus-like morphology.
• Figure 3: Stacked boundaries profiles used to construct a five-layer corneal model. Each interface curve defines one layer boundary, enabling independent control of the thickness profiles for the epithelium, Bowman’s layer, stroma, Descemet’s membrane, and endothelium; the upper boundary, air (propagation gap), and vitreous boundary are included for completeness.
• Figure 4: (a) Simulated cornea section anchored to clinically relevant scan widths (3--9 mm). The upper portion of panel (a) includes adapted material from Scientific Reports (DOI: 10.1038/s41598-021-93186-9) under CC BY 4.0heslinga2021corneal and was re-rendered and annotated. (b) Cropped multilayer corneal area from panel (a) and the final calibrated visualization, obtained by applying the aspect-ratio.
• Figure 5: A representative sample set produced by the proposed framework.