Optical coherence tomography image denoising using a generative adversarial network with speckle modulation.

Area of Science:

• Biomedical Imaging
• Optical Coherence Tomography
• Artificial Intelligence

Background:

• Speckle noise significantly degrades Optical Coherence Tomography (OCT) image quality.
• High-quality OCT images are crucial for accurate biomedical imaging and clinical diagnosis.

Purpose of the Study:

• To develop a novel deep learning model for effective OCT image denoising.
• To improve the quality of OCT images and videos by reducing speckle noise.

Main Methods:

• Developed a custom generative adversarial network (GAN), termed SM-GAN.
• Utilized a speckle-modulating OCT (SM-OCT) to generate low-speckle ground truth images.
• Trained and validated the SM-GAN model using 210,000 SM-OCT images.

Main Results:

• SM-GAN effectively reduced speckle noise in various OCT datasets, including retinal images, 3D finger scans, and dynamic OCT videos.
• The model maintained essential spatial and temporal resolutions during the denoising process.
• Performance surpassed traditional denoising methods and other deep learning approaches.

Conclusions:

• The proposed SM-GAN model offers a powerful solution for denoising OCT images and videos.
• This advancement has the potential to enhance diagnostic accuracy and broaden OCT applications.
• SM-GAN preserves critical image details, making it suitable for real-time biomedical imaging.