Reconstruction of Optical Coherence Tomography Images from Wavelength Space Using Deep Learning.

Area of Science:

• Biomedical Imaging
• Optical Coherence Tomography
• Deep Learning

Background:

• Conventional Fourier Domain Optical Coherence Tomography (FD-OCT) requires wavenumber domain resampling, increasing hardware and computational demands.
• OCT images are inherently affected by speckle noise due to low-coherence interferometry.

Purpose of the Study:

• To propose a computationally efficient Deep Learning (DL) approach for reconstructing speckle-reduced OCT images directly from the wavelength (λ) domain.
• To reduce the computational complexity associated with traditional FD-OCT methods.

Main Methods:

• Sequential application of two encoder-decoder Deep Learning networks: Spatial Domain Convolution Neural Network (SD-CNN) and Fourier Domain CNN (FD-CNN).
• SD-CNN reconstructs morphological structures and suppresses noise from degraded images derived from wavelength domain fringes.
• FD-CNN further optimizes image quality in the Fourier domain (FD).

Main Results:

• Demonstrated quantitative and visual efficacy in obtaining high-quality, speckle-reduced OCT images.
• Illustrated significant reduction in computational complexity compared to conventional methods.
• Successfully reconstructed OCT images directly from the wavelength domain, bypassing wavenumber domain resampling.

Conclusions:

• The proposed DL-based method offers a streamlined and computationally efficient alternative for OCT image reconstruction.
• This approach effectively reduces speckle noise and enhances image quality.
• The work provides a foundation for future advancements in OCT image processing and reconstruction.