Energy-guided learning approach to compressive FD-OCT.

Area of Science:

• Biomedical Imaging
• Optical Coherence Tomography
• Data Acquisition

Background:

• High-quality volumetric imaging using optical coherence tomography (OCT) generates large datasets, risking motion artifacts and exceeding radiation exposure limits.
• Sparse or compressive sampling strategies are crucial for recovering OCT signals from sub-Nyquist measurements, mitigating data acquisition challenges.

Purpose of the Study:

• To introduce a novel energy-guided compressive sensing (CS) approach for Fourier domain OCT (FD-OCT).
• To enhance image quality and reduce data acquisition volume in biological tissue imaging.

Main Methods:

• Developed an algorithm that learns an optimized sampling probability density function based on training data energy distribution.
• Implemented energy-guided sparse sampling for FD-OCT, contrasting it with conventional uniform random sampling.
• Applied the method to reconstruct retina images from sparse datasets.

Main Results:

• The energy-guided learning approach required 45% fewer samples compared to uniform CS for retina imaging.
• Achieved comparable reconstruction performance to conventional CS methods.
• Demonstrated significant reduction in data acquisition while preserving image quality.

Conclusions:

• The proposed energy-guided compressive sensing approach offers a significant advancement for OCT imaging.
• This method has the potential to substantially decrease data acquisition requirements without compromising image quality.
• Enables more efficient and safer volumetric imaging of biological tissues.