GARD: Gamma-based Anatomical Restoration and Denoising for Retinal OCT

2025-09-12

AI Summary

OCT images are severely degraded by multiplicative speckle noise, obscuring fine anatomical structures and posing challenges for simultaneous denoising and structural fidelity preservation. To address this, we propose an anatomy-aware diffusion-based denoising framework grounded in Gamma distribution modeling. First, the Gamma distribution is adopted to more accurately characterize the statistical properties of OCT speckle noise. Second, an anatomy fidelity constraint is incorporated into the DDIM sampling process to suppress high-frequency artifacts and enhance reconstruction of retinal layer boundaries and microstructures. Third, a preprocessing-guided inference acceleration mechanism is introduced. Evaluated on paired OCT datasets, our method achieves statistically significant improvements in PSNR, SSIM, and MSE over conventional filters and state-of-the-art deep learning models. Qualitative assessment confirms superior edge sharpness, inter-layer contrast, and preservation of clinically critical anatomical details.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Distributed Search
• Cognitive Modeling & Cognitive Systems: Neural Spike Coding

Application Category

• Security and Privacy: Data transparency and provenance
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Optical Coherence Tomography (OCT) is a vital imaging modality for diagnosing and monitoring retinal diseases. However, OCT images are inherently degraded by speckle noise, which obscures fine details and hinders accurate interpretation. While numerous denoising methods exist, many struggle to balance noise reduction with the preservation of crucial anatomical structures. This paper introduces GARD (Gamma-based Anatomical Restoration and Denoising), a novel deep learning approach for OCT image despeckling that leverages the strengths of diffusion probabilistic models. Unlike conventional diffusion models that assume Gaussian noise, GARD employs a Denoising Diffusion Gamma Model to more accurately reflect the statistical properties of speckle. Furthermore, we introduce a Noise-Reduced Fidelity Term that utilizes a pre-processed, less-noisy image to guide the denoising process. This crucial addition prevents the reintroduction of high-frequency noise. We accelerate the inference process by adapting the Denoising Diffusion Implicit Model framework to our Gamma-based model. Experiments on a dataset with paired noisy and less-noisy OCT B-scans demonstrate that GARD significantly outperforms traditional denoising methods and state-of-the-art deep learning models in terms of PSNR, SSIM, and MSE. Qualitative results confirm that GARD produces sharper edges and better preserves fine anatomical details.

Problem

• Reduces speckle noise in retinal OCT images
• Preserves fine anatomical structures during denoising
• Accurately models statistical properties of OCT noise

Innovation

• Denoising Diffusion Gamma Model
• Noise-Reduced Fidelity Term
• Denoising Diffusion Implicit Model