Three-dimensional coherent X-ray diffraction imaging via deep convolutional neural networks

Wu, Luhua; Yoo, Shinjae; Suzana, Ana F.; Assefa, Tadesse; Diao, Jiecheng; Harder, Ross; Cha, Wonsuk; Robinson, Ian

doi:10.1038/s41524-021-00644-z

Cited by 42 publications

(33 citation statements)

References 44 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Besides, the authors assume the object images are magnitude-only (no phase information), which is far from real situations. As for NNPhase [24], it aims to deal with the crystal data that the application is quite specific. Besides, LenslessNet [18] is initially designed for the phase retrieval task based on Fresnel diffraction patterns with lensless imaging systems.…”

Section: Phase-only Datasetsmentioning

confidence: 99%

“…Besides, the conditional generative adversarial network (GAN), a kind of data-driven generative method based on specific conditions [21], was adopted for reconstructing the phase information from Fourier intensity measurements [22]. Wu et al also proposed a CNN to reconstruct the crystal structures with coherent X-ray diffraction patterns [23,24]. The above methods adopt supervised learning that trains the networks with the paired ground-truth data.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

SiSPRNet: End-to-End Learning for Single-Shot Phase Retrieval

Ye¹,

Wang²,

Lun³

2022

Preprint

View full text Add to dashboard Cite

Traditional optimization algorithms have been developed to deal with the phase retrieval problem. However, multiple measurements with different random or non-random masks are needed for giving a satisfactory performance. This brings a burden to the implementation of the algorithms in practical systems. Even worse, expensive optical devices are required to implement the optical masks. Recently, deep learning, especially convolutional neural networks (CNN), has played important roles in various image reconstruction tasks. However, traditional CNN structure fails to reconstruct the original images from their Fourier measurements because of tremendous domain discrepancy. In this paper, we design a novel CNN structure, named SiPRNet, to recover a signal from a single Fourier intensity measurement. To effectively utilize the spectral information of the measurements, we propose a new Multi-Layer Perception block embedded with the dropout layer to extract the global representations. Two Up-sampling and Reconstruction blocks with self-attention are utilized to recover the signals from the extracted features. Extensive evaluations of the proposed model are performed using different testing datasets on both simulation and optical experimentation platforms. The results demonstrate that the proposed approach consistently outperforms other CNN-based and traditional optimizationbased methods in single-shot maskless phase retrieval. The source codes of the proposed method have been released on Github: https://github.com/Qiustander/SiPRNet.

show abstract

Section: Phase-only Datasetsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

SiSPRNet: End-to-End Learning for Single-Shot Phase Retrieval

Ye¹,

Wang²,

Lun³

2022

Preprint

View full text Add to dashboard Cite

show abstract

“…This can result in a total number of summation of the order of 10 9 for a pixelated imaging plane. Although this approach is not always prohibitive on modern accelerated Graphical Processing Unit (GPU) computing facilities or High Performance Computing (HPC) facilities, there are certain redundancies (outlined below) in the standard general approach that when removed, further and significantly increases the efficiency of the calculation and enable its use in more demanding applications such as training data for machine learned phase retrieval optimisation [19,20].…”

Section: Introductionmentioning

confidence: 99%

Simulation of Bragg coherent diffraction imaging

2022

View full text Add to dashboard Cite

The arrangement of atoms within a crystal and information on deviations from the ideal lattice is encoded in the diffraction pattern obtained from an appropriately conducted Bragg coherent diffraction imaging (BCDI) experiment. A foreknowledge of how specific displacements of atoms within the unit cell alter the BCDI diffraction pattern and the subsequent real-space image is often useful for interpretation and can provide valuable insight for materials design. Here we report on an atomistic approach to efficiently simulate BCDI diffraction patterns by factorising and eliminating certain redundancies in the conventional approach. Our method is able to reduce the computation time by several orders of magnitude without compromising the recovered phase information and therefore enables feasible atomistic simulations on nanoscale crystals with arbitrary lattice distortions.

show abstract

“…It has been demonstrated that a deep neural network-based method, which is a non-iterative end-to-end method, provides rapid results for phase retrieval in 2D and 3D coherent X-ray imaging [23][24][25][26]. Moreover, there have been rapid progresses for optical tomography [27,28], ghost imaging [29,30], face detection [31], growth stage detection [32], and low photon imaging [33].…”

Section: Introductionmentioning

confidence: 99%

Performance Evaluation of Deep Neural Network Model for Coherent X-ray Imaging

Kim

Messerschmidt

Graves

2022

View full text Add to dashboard Cite

We present a supervised deep neural network model for phase retrieval of coherent X-ray imaging and evaluate the performance. A supervised deep-learning-based approach requires a large amount of pre-training datasets. In most proposed models, the various experimental uncertainties are not considered when the input dataset, corresponding to the diffraction image in reciprocal space, is generated. We explore the performance of the deep neural network model, which is trained with an ideal quality of dataset, when it faces real-like corrupted diffraction images. We focus on three aspects of data qualities such as a detection dynamic range, a degree of coherence and noise level. The investigation shows that the deep neural network model is robust to a limited dynamic range and partially coherent X-ray illumination in comparison to the traditional phase retrieval, although it is more sensitive to the noise than the iteration-based method. This study suggests a baseline capability of the supervised deep neural network model for coherent X-ray imaging in preparation for the deployment to the laboratory where diffraction images are acquired.

show abstract

Three-dimensional coherent X-ray diffraction imaging via deep convolutional neural networks

Cited by 42 publications

References 44 publications

SiSPRNet: End-to-End Learning for Single-Shot Phase Retrieval

SiSPRNet: End-to-End Learning for Single-Shot Phase Retrieval

Simulation of Bragg coherent diffraction imaging

Performance Evaluation of Deep Neural Network Model for Coherent X-ray Imaging

Contact Info

Product

Resources

About