Related Experiment Video
Updated: May 5, 2026
05:52
Multi-Fiber Photometry to Record Neural Activity in Freely-Moving Animals
Published on: October 20, 2019
36.9K
Physics-constrained Mamba-UNet for enhanced-resolution single multimode fiber ghost imaging at low sampling rates
Optics Express
|May 4, 2026
Summary
We developed a novel Physics-Constrained Mamba-UNet for single multimode fiber (MMF) ghost imaging, enhancing endoscopic applications. This method achieves high-resolution reconstruction even at low sampling rates, overcoming mode coupling challenges.
Area of Science:
- Optics and Photonics
- Computational Imaging
- Biomedical Engineering
Background:
- Single multimode fiber (MMF) ghost imaging shows promise for endoscopy.
- Practical use is limited by mode coupling and ill-posed reconstruction at low sampling rates.
Purpose of the Study:
- To propose a high-fidelity reconstruction framework for MMF ghost imaging.
- To enhance resolution and robustness for endoscopic applications.
Main Methods:
- Physics-Constrained Mamba-UNet incorporating a visual state space model (VSSM).
- Integration of a differentiable forward physical observation model.
- Training as a physics-constrained inverse problem.
Main Results:
- Achieved 2.2-fold resolution enhancement and high structural similarity.
- Demonstrated superior performance at sampling rates as low as 5%.
- Framework shows improved robustness against mode-mixing noise.
Conclusions:
- The proposed framework offers a viable pathway for real-time, high-resolution fiber endoscopy.
- Physics-constrained deep learning effectively addresses challenges in MMF ghost imaging.
