Retinal Optical Coherence Tomography Image Denoising Using Modified Soft Thresholding Wavelet Transform

Optical Coherence Tomography (OCT) represents a non-invasive imaging modality capable of capturing high-resolution cross-sectional images of anatomical structures by scanning the tissue of interest in a transverse manner. Nevertheless, the inherent speckle noise present in OCT images considerably degrades their textural and sharpness qualities. Conventional wavelet-based modified soft thresholding methods have been employed to preserve pertinent information in denoising OCT images, but their performance remains contingent upon hyperparameter tuning. In this study, we introduce a Particle Swarm Optimization (PSO)-based optimized Wavelet Threshold (WT) method for OCT image denoising. By automating the process of determining hyperparameter values dependent on image quality, PSO streamlines the denoising process. The optimization problem's fitness function is defined by the Peak Signal-to-Noise Ratio (PSNR) parameter. To evaluate the WT-PSO algorithm, we utilized performance metrics such as Mean Square Error (MSE), PSNR, Structural Similarity Index Metrics (SSIM), and Contrast-to-Noise Ratio (CNR) on a publicly available dataset comprising 17 retinal OCT images. The proposed denoising approach demonstrates comparable results to those obtained by manual iterative or trial methods, delivering marginal improvements in performance parameters and image quality. Moreover, our method outperforms traditional wavelet-based state-of-the-art techniques for denoising OCT images, highlighting its potential for widespread application in the field.