Enhancing Optical Coherence Tomography Images Of Central Serous Retinopathy Using EAC-NLM Algorithm: A Quantitative Evaluation

Abstract

Introduction: The intricacies involved in retinal imaging make it difficult to accurately diagnose and track Central Serous Retinopathy (CSR). This disorder is characterized by anomalies in the layers of retina that make up the retina as well as fluid leaking that usually happens around the macula.
Objectives: Pressure from the condition builds up inside the layers of the retina, causing the retinal walls to separate and impede vision. This study explores the application of the proposed Enhanced Adaptive Contrast Non-Local Means (EAC-NLM) algorithm to enhance Optical Coherence Tomography (OCT) images of CSR. The study utilized OCT images from Rajiv Gandhi Government General Hospital, Chennai, Tamil Nadu, acquired using a Spectralis OCT scanner.
Methods: A dataset comprising macula-centered SD-OCT scans of 50 eyes for testing and 10 images for training was processed. Each OCT volume image had a resolution of 512 × 128 × 1024 voxels with voxel dimensions of 10.90 × 45.00 × 2.00 μm³.
Results: Quantitative evaluation using image quality metrics further substantiates the effectiveness of EAC-NLM. The denoised images of proposed EAC-NLM show high structural similarity (SSIM: 0.9850), excellent fidelity (PSNR: 45.0000 dB), and minimal error (MSE: 2.500e-05) compared to the original, validating the algorithm's effectiveness in enhancing OCT images for CSR diagnosis.
Conclusions: This study explores the application of the proposed Enhanced Adaptive Contrast Non-Local Means (EAC-NLM) algorithm to enhance Optical Coherence Tomography (OCT) images of CSR.