Related Experiment Video
Updated: May 10, 2025
10:16
Digital Inline Holographic Microscopy DIHM of Weakly-scattering Subjects
Published on: February 8, 2014
12.2K
Adaptive Vectorial Restoration from Dynamic Speckle Patterns Through Biological Scattering Media Based on Deep
Yu-Chen Chen1, Shi-Xuan Mi1, Ya-Ping Tian1
1Key Laboratory of Optical Field Manipulation of Zhejiang Province, Department of Physics, Zhejiang Sci-Tech University, Hangzhou 310018, China.
Sensors (Basel, Switzerland)
|April 28, 2025
Summary
This study introduces a deep learning method for advanced biomedical imaging. It uses vector optical fields to reconstruct images from dynamic biological tissues, improving accuracy and robustness in scattering environments.
Area of Science:
- Biomedical Optics
- Deep Learning in Imaging
- Vector Optical Fields
Background:
- Vector optical fields offer potential for non-invasive imaging of biological tissues.
- Dynamic and anisotropic biological tissues present challenges for light propagation and imaging due to scattering.
- Existing methods struggle with accurate reconstruction in complex scattering media.
Purpose of the Study:
- To develop a robust deep learning-based method for polarization-resolved imaging reconstruction.
- To address challenges in imaging dynamic and anisotropic biological tissues using vector optical fields.
- To enhance imaging accuracy and efficiency in scattering environments.
Main Methods:
- A novel deep learning approach utilizing polarization-resolved restoration.
- Integration of Convolutional Neural Networks (CNNs) and Transformer architectures in a hybrid network.
- Leveraging two orthogonal polarization components of vector optical fields for enhanced information capture.
Main Results:
- Demonstrated excellent robustness and generalization for reconstructing images from dynamic speckle patterns.
- Successfully achieved efficient and accurate imaging reconstruction through anisotropic and time-varying scattering media.
- Validated the model's capability in capturing both local and global features of speckle images.
Conclusions:
- The proposed method provides an efficient solution for scattering imaging of dynamic anisotropic biological tissues.
- This work advances the application of vector optical fields in dynamic scattering environments by integrating deep learning.
- The hybrid network model shows significant potential for future biomedical imaging applications.
