SUMMARY:
UCLA researchers in the Department of Electrical & Computer Engineering have developed an innovative recurrent neural network algorithm that improves image quality and increases the speed of holographic image reconstruction.
BACKGROUND:
Tissue biopsies are a commonly used tool for disease diagnosis. Pathologists have been transitioning into more data-driven approaches for analyzing pathology samples. Digital holographic microscopy is a powerful tool to obtain three-dimensional and complex (phase and amplitude) transmittance properties about biological samples with minimal sample preparation and without expensive microscopes. This technology works by acquiring images of a sample at various focus heights and then computationally reconstructing the complex field of samples in three-dimensional space. However, conventionally complicated algorithms must be employed to reconstruct sample phase and amplitude information from intensity-only measurement by photoelectric sensors. The standard computational tools used tend to be iterative and slow in nature. There is a clear need for algorithms that are capable of quickly recovering high degrees of spatial resolution. While recent research has shown that deep-learning reconstruction algorithms outperform conventional iterative techniques, there is a need to further develop these algorithms to improve image quality and reduce processing time.
INNOVATION:
UCLA researchers have developed a new recurrent neural network-based (RNN) holographic reconstruction algorithm that can efficiently reconstruct the phase and amplitude of an imaged sample. The algorithm improves image quality reconstruction by 40% at a processing rate of 15-fold compared to other deep-learning based iterative reconstruction processes. In addition, the invention reduces the complexity of the reconstruction process by removing the need for free-space back-propagation (FSP) step in existing deep learning-based holographic reconstruction algorithms. The algorithm also allows for a greater depth of field due its integrated auto-focusing feature. This innovation provides an invaluable tool for utilization in various coherent imaging applications.
POTENTIAL APPLICATIONS:
- Digital Holographic Microscopy
- Other coherent imaging modalities
- 3D image reconstruction for disease diagnostics
ADVANTAGES:
- High reconstruction quality
- High reconstruction speed
- Extended depth of field
- Autofocus capability
- Low complexity
DEVELOPMENT-TO-DATE:
First successful demonstration (first actual reduction to practice) has been accomplished.
RELATED PAPERS:
Huang, L.; Liu, T.; Yang, X.; Luo, Y.; Rivenson, Y.; Ozcan, A. Holographic Image Reconstruction with Phase Recovery and Autofocusing Using Recurrent Neural Networks. ACS Photonics 2021, 8 (6), 1763–1774. https://doi.org/10.1021/acsphotonics.1c00337.
Wu, Y.; Rivenson, Y.; Zhang, Y.; Wei, Z.; Günaydin, H.; Lin, X.; Ozcan, A. Extended Depth-of-Field in Holographic Imaging Using Deep-Learning-Based Autofocusing and Phase Recovery. Optica 2018, 5 (6), 704. https://doi.org/10.1364/optica.5.000704.