Volumetric non-local-means based speckle reduction for optical coherence tomography

Area of Science:

• Medical Imaging
• Biomedical Optics
• Image Processing

Background:

• Speckle noise degrades the quality of Optical Coherence Tomography (OCT) images.
• Existing despeckling methods may compromise image resolution or struggle with motion artifacts.

Purpose of the Study:

• To introduce TNode, a novel tomographic non-local-means based despeckling technique for OCT.
• To evaluate TNode's effectiveness in reducing speckle noise and preserving image details in 3D OCT volumes.

Main Methods:

• Developed a tomographic non-local-means algorithm (TNode) utilizing a 3D similarity window for speckle weighting.
• Implemented two versions: one with a 2D search window for motion artifact tolerance and another with a 3D window for motion-free data.
• Applied TNode to phantom, ophthalmic, and esophageal OCT datasets.

Main Results:

• TNode achieved effective speckle reduction comparable to B-scan compounding and out-of-plane averaging.
• The technique preserved isotropic resolution, including speckle-sized structures.
• Small retinal vessels were clearly visualized in en-face and cross-sectional views.
• Despeckled esophageal volumes showed improved image quality and detail, even with motion.

Conclusions:

• TNode offers robust speckle reduction for 3D OCT volumes.
• The method preserves fine image details and isotropic resolution without requiring dictionaries or segmentation.
• TNode is broadly applicable to various OCT datasets, including those with motion artifacts, enhancing diagnostic capabilities.