- Nuclear & Particle Physics
- Isotope Research & Production
- RIKEN BNL Research Center
Working at CFN
- Arrival & Departure
- Reports & Publications
- Acknowledging Use of CFN Facilities
- Data Management
- The Guide to Brookhaven
- Operations Plan
- Experimental Safety Reviews (ESR)
- COSA Training
- Hours of Operation
- Laser System Qualification
- Transport of Hazardous Materials
- Vendor Registration (PDF)
- News & Events
Machine Learning Accelerates High-Resolution X-ray Imaging
January 31, 2020
What is the scientific achievement?
A collaborative team from CFN, NSLS-II, and Stony Brook University created a machine-learning algorithm based on a convolutional neural network that accelerates the process of imaging materials with coherent X-rays. This imaging method, called X-ray ‘ptychography,’ is a powerful, high-resolution technique that typically requires long experimental and computational time. The machine-learning algorithm accelerates ptychographic imaging by around 90% based on simulations compared to conventional methods.
Why does this achievement matter?
The speed provided by this new, machine learning-based method makes possible the use of X-ray ptychography for high-resolution studies of beam sensitive materials, and to image in-situ dynamics of nanomaterials in different environments.
What are the details?
Ptychography is a coherent diffractive imaging method that captures multiple diffraction patterns of a sample with a set of shifted localized illuminations (“probes”). The reconstruction problem, known as “phase retrieval,” is typically solved by iterative algorithms. In this paper, we propose PtychoNet, a deep learning based method to perform phase retrieval for ptychography in a non-iterative manner. We devise a generative network to encode a full ptychography scan, reverse the diffraction at each scanning point and compute the amplitude and phase of the object. We demonstrate successful reconstructions using PtychoNet as well as recovering fine features in the case of extreme sparse scanning where conventional iterative methods fail to give recognizable features.
The CFN Theory and Computation Facility was used for this work.
Z. Guan, E.H.R. Tsai, X. Huang, K.G. Yager, H. Qin, PtychoNet: Fast and High Quality Phase Retrieval for Ptychography, British Machine Vision Conference 1172 (2019)
BNL Newsroom: CFN Staff Spotlight: Esther Tsai Brings Expertise in X-ray Imaging
Acknowledgement of Support
This work was supported by NSF IIS-1715985 and NSF IIS-1812606.
This research used resources of the Center for Functional Nanomaterials and the National Synchrotron Light Source II, both of which are U.S. Department of Energy (DOE) Office of Science Facilities operated for the DOE Office of Science by Brookhaven National Laboratory under Contract No. DE-SC0012704.
2020-17091 | INT/EXT | Newsroom