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

Abstract

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.

Towards fast ptychography image reconstruction of EUV masks by deep neural networks

“…In the last few years several works have been published on the topic of phase retrieval using deep neural networks (DNN), comprising both the single intensity phase retrieval problem and ptychography, e.g. [9][10][11][12] In these approaches, a DNN architecture is trained to learn the mapping from the diffraction data to the real space (complex-valued) object. Once trained, the network is used at inference time, to estimate an unseen object given a set of diffraction patterns generated from i t. While training these networks is a long process, inference is usually very fast, hence making this approach suitable for fast and high-quality imaging of a given object from its diffraction data.…”