A deep-learning-based approach for noise reduction in high-speed optical coherence Doppler tomography.

Area of Science:

• Biomedical Optics
• Medical Imaging
• Machine Learning in Medicine

Background:

• Optical Coherence Doppler Tomography (ODT) offers high contrast, capillary-level resolution, and flow speed quantification.
• A trade-off exists between ODT image signal-to-noise ratio and A-scan sampling density, limiting imaging speed and clinical use.

Purpose of the Study:

• To develop a deep-learning-based approach for accelerating ODT imaging by suppressing phase noise.
• To overcome limitations of small paired training datasets using generative adversarial networks.

Main Methods:

• A generative adversarial network (GAN) was used to learn Doppler phase noise distribution and generate synthetic data.
• A 3D convolutional neural network (CNN) was trained for image denoising using the synthetic data.
• The deep learning approach was applied to suppress noise in low-sampling density ODT images.

Main Results:

• The proposed deep learning method significantly reduced noise in ODT images.
• Image details were well-preserved, outperforming traditional denoising techniques.
• High-speed ODT imaging was achieved with low A-scan sampling density.

Conclusions:

• Deep learning effectively suppresses phase noise in ODT imaging.
• This approach enables high-speed ODT imaging without compromising image quality.
• The method holds promise for expanding clinical applications of ODT.