Deep Microlocal Reconstruction
By H. Andrade-Loarca, G. Kutyniok, O. Öktem, P. Petersen
License
This code is released under the MIT License (refer to the LICENSE file for details).
Contents
Introduction
This repository contains the entire pipline (including data preprocessing, training, testing, evaluation and visualization) for Deep Microlocal Reconstruction for Limited-Angle Tomography.
The algorithm is based on a recently developed digital wavefront set extractor DeNSE as well as the well-known microlocal canonical relation for the Radon transform. We use the wavefront set information about x-ray data to improve the reconstruction by requiring that the underlying neural networks simultaneously extract the correct ground truth wavefront set and groundtruth image. As a necessary theoretical step, we identify the digital microlocal canonical relations for deep convolutional residual neural networks. We find strong numerical evidence for the effectiveness of this approach
Citation
If you find Deep Microlocal Reconstruction for Limited-Angle Tomography useful in your research, please consider to cite the following papers:
@inproceedings{andrade2021deepmicrorecon,
title={Deep Microlocal Reconstruction for Limited-Angle Tomography},
author={Andrade-Loarca, Hector, Kutyiniok, Gitta, Öktem, Ozan, Petersen, Philipp},
booktitle={arXiv preprint: arXiv:2108.05732 },
year={2021}
}Installation
You can install all the dependencies by using the conda local enviroment file.
conda env create -f environment.yaml
conda activate deepmicroUsage
The different experiments presented in the preprint are here organized in different folders. The folder Joint_CT_WFset_inpaint contains the experiments corresponding to the joint reconstruction. The folder Microlocal_NN contains the code corresponding to the microcanonical relation of the learned primal-dual reconstruction. The folder Real_phantoms contains the code necessary to generate the realistic phantoms data set that was used to train the reconstruction model. Finally, the folder WF_inpaint contains the experiments corresponding to the wavefront set inpainting.
Contact
Questions can also be left as issues in the repository. We will be happy to answer them.