2019
DOI: 10.3390/app9122445
|View full text |Cite
|
Sign up to set email alerts
|

On-the-Fly Machine Learning for Improving Image Resolution in Tomography

Abstract: In tomography, the resolution of the reconstructed 3D volume is inherently limited by the pixel resolution of the detector and optical phenomena. Machine learning has demonstrated powerful capabilities for super-resolution in several imaging applications. Such methods typically rely on the availability of high-quality training data for a series of similar objects. In many applications of tomography, existing machine learning methods cannot be used because scanning such a series of similar objects is either imp… Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

1
9
0

Year Published

2019
2019
2023
2023

Publication Types

Select...
6
1

Relationship

3
4

Authors

Journals

citations
Cited by 9 publications
(10 citation statements)
references
References 38 publications
1
9
0
Order By: Relevance
“…The figure shows that U-Net and DnCNN started to fit the noise, whereas the PSNR of the MS-D network continued to increase. This matches earlier results on overfitting [10], [39], [54]. If the training dataset had been larger, these effects could have been less pronounced.…”
Section: F Hyper-parameterssupporting
confidence: 90%
“…The figure shows that U-Net and DnCNN started to fit the noise, whereas the PSNR of the MS-D network continued to increase. This matches earlier results on overfitting [10], [39], [54]. If the training dataset had been larger, these effects could have been less pronounced.…”
Section: F Hyper-parameterssupporting
confidence: 90%
“…Another major advantage of MS‐D networks over U‐Net and ResNet is the use of dilated convolutional kernels instead of standard convolutional kernels. This allows MS‐D networks to learn which combinations of dilations are most suited to solve the task at hand and offers the unique possibility to use the same MS‐D network architecture for a broad range of different applications such as segmenting organelles in microscopic cell images, image denoising and improving the resolution of tomographic reconstructions . Finally, all layers of an MS‐D network are interconnected using the same set of standard operations [see Section 2, Eqs.…”
Section: Discussionmentioning
confidence: 99%
“…For performance reasons, the simulated phantoms are generated through C++ using Cython [ 65 ]. For the evaluation of U-nets, we took the PyTorch [ 66 ] implementation used in [ 67 ]. The MSD-nets are implemented using the package published with [ 26 ].…”
Section: Appendix A1 Data Generationmentioning
confidence: 99%
“…For the evaluation of U-nets, we took the PyTorch [ 66 ] implementation used in [ 67 ]. The MSD-nets are implemented using the package published with [ 26 ].…”
Section: Appendix A1 Data Generationmentioning
confidence: 99%