Smooth Gyrated Texel Quadrivium Network for 3D Retinal OCT Image Compression

Area of Science:

• Medical Imaging
• Computer Vision
• Biomedical Engineering

Background:

• Retinal optical coherence tomography (OCT) image compression requires effective pre-processing for noise reduction.
• Existing denoising methods struggle with signal-dependent noise and hybridized low-contrast residual noise (HLRN).
• Current segmentation techniques often overlook retinal nerve deviation, impacting image interpretability.

Purpose of the Study:

• To propose a novel network, the Smooth Gyrated Texel Quadrivium Network (SGTQN), for advanced OCT image denoising and information gathering.
• To develop an improved segmentation method that incorporates retinal nerve deviation for self-explanatory images.
• To enhance OCT image compression quality through uniform compression techniques.

Main Methods:

• The SGTQN utilizes an Additive Ascombe Smooth Sifter to convert Poisson noise to Gaussian noise and remove HLRN.
• An Improvised Gyrated Alexa Net with a Standardized Gyrated Layer is employed for segmentation, considering deviation values.
• A Texel Quadrivium Convolutional Network modifies the pooling layer to a Texel Quadrivium Layer for uniform compression.

Main Results:

• The proposed model effectively reduces both signal-dependent and HLRN in OCT images.
• The segmentation method generates self-explanatory images by incorporating retinal nerve deviation.
• High-quality image compression is achieved with uniform compression and adjuvant vector coordinates.
• The model demonstrates a high accuracy of 95% and a low mean square error of 0.02.

Conclusions:

• The SGTQN offers a robust solution for denoising and enhancing information extraction in retinal OCT images.
• The integrated segmentation approach improves image interpretability by considering nerve deviation.
• The novel compression technique yields high-resolution, high-quality compressed images, advancing OCT analysis.