Super-resolution reconstruction for FDOCT based on order calculated MUSIC spectral estimation

Area of Science:

• Optical Engineering
• Biomedical Imaging
• Signal Processing

Background:

• Frequency-domain optical coherence tomography (FD-OCT) typically uses Fourier transforms for depth profiling.
• This method is limited to integer multiples of coherence length, preventing non-integer component resolution.
• Existing methods struggle with precise depth localization and noise reduction.

Purpose of the Study:

• To introduce a super-resolution image reconstruction algorithm for FD-OCT.
• To overcome the resolution limitations of Fourier transform-based methods.
• To improve axial depth intensity profile accuracy and noise suppression.

Main Methods:

• Developed a super-resolution algorithm utilizing Multiple Signal Classification (MUSIC) spectral estimation.
• Analyzed eigenvalue intensity statistics of the autocorrelation matrix to determine spectral estimation order.
• Implemented noise suppression by aligning spectral estimation order with the noise model.

Main Results:

• Achieved a resolution of 3.2μm, a 4.7x improvement over Fourier-based reconstruction.
• Demonstrated significant background noise suppression.
• Validated performance on wedge-shaped air gaps and fingertip skin samples.
• Showcased accurate optimal order selection for enhanced image reconstruction.

Conclusions:

• The MUSIC-based super-resolution algorithm significantly enhances FD-OCT axial resolution and image quality.
• The method effectively suppresses noise and precisely localizes depth profiles.
• This approach offers a substantial advancement over conventional Fourier transform techniques in OCT imaging.