N2NSR-OCT: Simultaneous denoising and super-resolution in optical coherence tomography images using semisupervised...

Area of Science:

• Biomedical Imaging
• Medical Technology
• Artificial Intelligence in Healthcare

Background:

• Optical coherence tomography (OCT) is a valuable noninvasive clinical imaging technique.
• OCT image quality is often degraded by speckle noise and low resolution.
• Improving OCT image quality is crucial for accurate clinical diagnosis.

Purpose of the Study:

• To develop a learning-based method for recovering high-quality OCT images from low-quality inputs.
• To simultaneously denoise and super-resolve OCT images within short scanning times.
• To enhance the signal-to-noise ratio (SNR) and resolution of OCT scans.

Main Methods:

• Proposed a semisupervised learning approach named N2NSR-OCT.
• Utilized up- and down-sampling networks (U-Net (Semi) and DBPN (Semi)).
• Trained models with 2× and 4× upscale factors for varying down-sampling rates.

Main Results:

• Achieved results comparable to supervised learning methods.
• Outperformed existing state-of-the-art methods in image quality.
• Successfully maintained subtle fine retinal structures in enhanced OCT images.

Conclusions:

• N2NSR-OCT offers an effective semisupervised solution for high-quality OCT image reconstruction.
• The method improves both denoising and super-resolution simultaneously.
• This approach has the potential to advance clinical OCT imaging by providing clearer, more detailed images.