Super-Resolution Optical Coherence Tomography Using Diffusion Model-Based Plug-and-Play Priors

2025-05-20

AI Summary

Sparse B-scan sampling in high-speed optical coherence tomography (OCT) degrades corneal image resolution and exacerbates noise, limiting clinical utility. Method: We propose the first plug-and-play super-resolution framework embedding a diffusion model, jointly modeling the sparse measurement inverse problem and data-driven priors. Our approach employs Markov Chain Monte Carlo (MCMC) posterior sampling for interpretable, high-fidelity reconstruction; leverages real-world high-speed undersampled training pairs—bypassing unrealistic synthetic downsampling assumptions; and integrates deep upsampling preprocessing to enhance convergence and stability. Results: Evaluated on both *in vivo* human and *ex vivo* fish corneal OCT data, our method significantly outperforms the 2D U-Net baseline, yielding sharper structural details and superior noise suppression. It establishes a new paradigm for clinical high-frame-rate, high-fidelity OCT imaging.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Sampling/Simulation-based Search
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Security and Privacy: Data transparency and provenance
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

We propose an OCT super-resolution framework based on a plug-and-play diffusion model (PnP-DM) to reconstruct high-quality images from sparse measurements (OCT B-mode corneal images). Our method formulates reconstruction as an inverse problem, combining a diffusion prior with Markov chain Monte Carlo sampling for efficient posterior inference. We collect high-speed under-sampled B-mode corneal images and apply a deep learning-based up-sampling pipeline to build realistic training pairs. Evaluations on in vivo and ex vivo fish-eye corneal models show that PnP-DM outperforms conventional 2D-UNet baselines, producing sharper structures and better noise suppression. This approach advances high-fidelity OCT imaging in high-speed acquisition for clinical applications.

Problem

• Reconstruct high-quality OCT images from sparse measurements
• Combine diffusion prior with MCMC for efficient posterior inference
• Improve OCT imaging fidelity for high-speed clinical applications

Innovation

• Uses diffusion model-based plug-and-play priors
• Combines diffusion prior with MCMC sampling
• Applies deep learning-based up-sampling pipeline