Optical tomography in a single camera frame using fringe-encoded deep-learning full-field OCT.

Area of Science:

• Biomedical Optics
• Medical Imaging
• Deep Learning Applications

Background:

• Optical coherence tomography (OCT) is crucial for in vivo imaging, offering high resolution, field-of-view, and working distance.
• Conventional OCT image reconstruction requires multiple phases, limiting imaging speed and increasing susceptibility to motion artifacts.

Purpose of the Study:

• To demonstrate a novel en face tomography method using a single camera frame.
• To enhance imaging speed and reduce artifacts in OCT for in vivo applications.
• To enable low-cost OCT systems using readily available cameras.

Main Methods:

• Encoding high-frequency fringe patterns into the sample layer using low-coherence interferometry.
• Utilizing a high-pass filter enhanced by deep learning networks (U-Net/Pix2Pix) for efficient pattern extraction.
• Validating the approach through in vivo imaging of human subjects' corneas.

Main Results:

• Achieved a 10-fold improvement in imaging speed compared to conventional OCT.
• Significantly reduced phase errors and motion-related artifacts in in vivo OCT.
• Demonstrated the feasibility of low-cost tomography with consumer-grade cameras.

Conclusions:

• A single-frame en face tomography approach is sufficient for high-quality OCT imaging.
• Deep learning significantly enhances fringe pattern extraction for rapid tomographic reconstruction.
• This method offers a cost-effective and faster alternative for in vivo OCT, with potential for broad applications.