Ultra-low sampling fiber imaging system based on physical-enhanced IPSR.

Area of Science:

• Optical imaging
• Biomedical engineering
• Computational imaging

Background:

• Multimode fiber (MMF) imaging offers ultra-fine diameters for minimally invasive endoscopy.
• Low sampling rates in MMF imaging cause significant fidelity loss, hindering real-time monitoring and increasing phototoxicity.
• Traditional compressed sensing methods struggle with severe information loss under extreme sampling conditions.

Purpose of the Study:

• To introduce a novel imaging framework, physical-enhanced IPSR (Phys-IPSR), for high-fidelity MMF imaging.
• To improve information encoding efficiency in MMF imaging systems operating at extremely low sampling rates.
• To establish a foundation for clinically compatible endoscopic systems.

Main Methods:

• Implemented a deterministic front-end screening strategy to minimize mutual coherence.
• Utilized a dual-stage network to retrieve high-frequency features from sparse data.
• Developed the Phys-IPSR framework for enhanced information recovery.

Main Results:

• Achieved high-resolution reconstruction (198x198 pixels) at an extremely low sampling rate (0.77%).
• Demonstrated a fidelity improvement exceeding 20 dB over traditional compressed sensing methods (PSNR: 36.46 dB, SSIM: 0.8682).
• Exhibited robustness within a 30 mm axial defocus range and rapid single-frame processing (0.047 seconds).

Conclusions:

• Phys-IPSR significantly enhances information encoding efficiency for MMF imaging under extreme sampling.
• The framework provides a design paradigm for high-fidelity fiber imaging, crucial for next-generation endoscopic systems.
• Phys-IPSR lays the groundwork for future clinically applicable endoscopic solutions.