Toward Ground-Truth Optical Coherence Tomography via Three-Dimensional Unsupervised Deep Learning Processing and...

Area of Science:

• Biomedical Imaging
• Artificial Intelligence in Medicine
• Optical Engineering

Background:

• Optical coherence tomography (OCT) provides high-resolution 3D imaging crucial for biomedical applications.
• Coherence speckle noise significantly degrades OCT image quality and limits its utility.
• Existing methods struggle to effectively remove speckle while preserving image details.

Purpose of the Study:

• To develop a novel speckle-free OCT imaging strategy.
• To leverage unsupervised 3D deep learning for noise reduction in OCT.
• To improve the performance and expand the applications of OCT imaging.

Main Methods:

• Introduction of toward-ground-truth OCT (tGT-OCT), a speckle-free OCT imaging strategy.
• Utilization of an unsupervised 3D deep-learning network with 3D convolutions.
• Training the network on random 3D volumetric data to differentiate speckle from true structures.

Main Results:

• tGT-OCT successfully distinguishes and separates speckle from real structures in 3D imaging.
• The method effectively reduces speckle noise, enhancing visibility of subtle structures.
• Spatial resolution is preserved, and imaging quality is significantly improved across diverse samples.

Conclusions:

• tGT-OCT offers a high-quality, speckle-free 3D OCT imaging solution.
• This approach surpasses previous state-of-the-art methods in noise reduction and structural preservation.
• The developed technique holds promise for advancing biomedical OCT applications.