Dark-field microscopy is a powerful technique for enhancing the imaging resolution and contrast of small unstained samples. In this study, we report a method based on end-to-end convolutional neural network to reconstruct high-resolution dark-field images from low-resolution bright-field images. The relation between bright- and dark-field which was difficult to deduce theoretically can be obtained by training the corresponding network. The training data, namely the matched bright- and dark-field images of the same object view, are simultaneously obtained by a special designed multiplexed image system. Since the image registration work which is the key step in data preparation is not needed, the manual error can be largely avoided. After training, a high-resolution numerical dark-field image is generated from a conventional bright-field image as the input of this network. We validated the method by the resolution test target and quantitative analysis of the reconstructed numerical dark-field images of biological tissues. The experimental results show that the proposed learning-based method can realize the conversion from bright-field image to dark-field image, so that can efficiently achieve high-resolution numerical dark-field imaging. The proposed network is universal for different kinds of samples. In addition, we also verify that the proposed method has good anti-noise performance and is not affected by the unstable factors caused by experiment setup.
%0 Journal Article
%1 Meng:20
%A Meng, Zhang
%A Ding, Liqi
%A Feng, Shaotong
%A Xing, FangJian
%A Nie, Shouping
%A Ma, Jun
%A Pedrini, Giancarlo
%A Yuan, Caojin
%D 2020
%I OSA
%J Opt. Express
%K ito kom reviewed
%N 23
%P 34266--34278
%R 10.1364/OE.401786
%T Numerical dark-field imaging using deep-learning
%U http://www.opticsexpress.org/abstract.cfm?URI=oe-28-23-34266
%V 28
%X Dark-field microscopy is a powerful technique for enhancing the imaging resolution and contrast of small unstained samples. In this study, we report a method based on end-to-end convolutional neural network to reconstruct high-resolution dark-field images from low-resolution bright-field images. The relation between bright- and dark-field which was difficult to deduce theoretically can be obtained by training the corresponding network. The training data, namely the matched bright- and dark-field images of the same object view, are simultaneously obtained by a special designed multiplexed image system. Since the image registration work which is the key step in data preparation is not needed, the manual error can be largely avoided. After training, a high-resolution numerical dark-field image is generated from a conventional bright-field image as the input of this network. We validated the method by the resolution test target and quantitative analysis of the reconstructed numerical dark-field images of biological tissues. The experimental results show that the proposed learning-based method can realize the conversion from bright-field image to dark-field image, so that can efficiently achieve high-resolution numerical dark-field imaging. The proposed network is universal for different kinds of samples. In addition, we also verify that the proposed method has good anti-noise performance and is not affected by the unstable factors caused by experiment setup.
@article{Meng:20,
abstract = {Dark-field microscopy is a powerful technique for enhancing the imaging resolution and contrast of small unstained samples. In this study, we report a method based on end-to-end convolutional neural network to reconstruct high-resolution dark-field images from low-resolution bright-field images. The relation between bright- and dark-field which was difficult to deduce theoretically can be obtained by training the corresponding network. The training data, namely the matched bright- and dark-field images of the same object view, are simultaneously obtained by a special designed multiplexed image system. Since the image registration work which is the key step in data preparation is not needed, the manual error can be largely avoided. After training, a high-resolution numerical dark-field image is generated from a conventional bright-field image as the input of this network. We validated the method by the resolution test target and quantitative analysis of the reconstructed numerical dark-field images of biological tissues. The experimental results show that the proposed learning-based method can realize the conversion from bright-field image to dark-field image, so that can efficiently achieve high-resolution numerical dark-field imaging. The proposed network is universal for different kinds of samples. In addition, we also verify that the proposed method has good anti-noise performance and is not affected by the unstable factors caused by experiment setup.},
added-at = {2021-02-25T15:07:46.000+0100},
author = {Meng, Zhang and Ding, Liqi and Feng, Shaotong and Xing, FangJian and Nie, Shouping and Ma, Jun and Pedrini, Giancarlo and Yuan, Caojin},
biburl = {https://puma.ub.uni-stuttgart.de/bibtex/27b9137c873c2a6896fab3092610475b6/vogelfrau},
doi = {10.1364/OE.401786},
interhash = {fe25bf64913aff49f1f4570edb77ccf1},
intrahash = {7b9137c873c2a6896fab3092610475b6},
journal = {Opt. Express},
keywords = {ito kom reviewed},
month = nov,
number = 23,
pages = {34266--34278},
publisher = {OSA},
timestamp = {2021-02-25T14:07:46.000+0100},
title = {Numerical dark-field imaging using deep-learning},
url = {http://www.opticsexpress.org/abstract.cfm?URI=oe-28-23-34266},
volume = 28,
year = 2020
}