HistoReg

This repository describes a framework [https://arxiv.org/abs/1904.11929] for automated registration of variably-stained digitized histology slices from various anatomical sites, based on a greedy diffeomorphic registration tool.

The dataset provided for the quantitative evaluation of the proposed approach was provided by the Automatic Non-rigid Histological Image Registration (ANHIR) challenge [1-2]. This challenge was a part of the IEEE International Symposium on Biomedical Imaging (ISBI) 2019 conference, see the workshop slides and used evaluation framework BIRL.

Citation

If you use HistoReg, please cite the following article:

@article{app11041892,
  author = {Venet, Ludovic and Pati, Sarthak and Feldman, Michael D. and Nasrallah, MacLean P. and Yushkevich, Paul and Bakas, Spyridon},
  title = {Accurate and Robust Alignment of Differently Stained Histologic Images Based on Greedy Diffeomorphic Registration},
  journal = {Applied Sciences},
  volume = {11},
  year = {2021},
  number = {4},
  article-number = {1892},
  url = {https://doi.org/10.3390/app11041892},
  issn = {2076-3417},
  doi = {10.3390/app11041892}
}

Supporting Grant

This work is partly supported by the National Institutes of Health under grant award numbers NIH/NCI/ITCR:U24-CA189523, NIH/NIBIB:R01EB017255, NIH/NIA:R01AG056014, and NIH/NIA:P30AG010124.

Disclaimer

• The software has been designed for research purposes only and has neither been reviewed nor approved for clinical use by the Food and Drug Administration (FDA) or by any other federal/state agency.
• This code (excluding dependent libraries) is governed by the license provided in here unless otherwise specified.

Build Requirements

• CMake: 2.8.12+
• C++ compiler: tested on GCC 4.9.2 and 7.4.0; MSVC 2015 on Windows

Download and Compile

By downloading HistoReg, you agree to our license.

Windows

Showing an example set of commands to download and compile HistoReg on a Windows machine:

git clone https://github.com/CBICA/HistoReg.git HistoReg
cd HistoReg
git submodule init 
git submodule update
mkdir bin
cd bin
cmake -A x64 .. # this will build ITK as HistoReg needs
cmake --build . --target ALL_BUILD
## wait for ITK to finish building
cmake -DCMAKE_INSTALL_PREFIX=${path_to_where_you_want_to_install} -A x64 .. # this is for HistoReg
# please ensure that install directory is empty and in a location different from source or bin folders
cmake --build . --target ALL_BUILD
cmake --build . --target INSTALL

Linux

On Linux or machines using make as the generator, please follow these guidelines:

git clone https://github.com/CBICA/HistoReg.git HistoReg
cd HistoReg
git submodule init 
git submodule update
mkdir bin
cd bin
cmake .. # this will build ITK as HistoReg needs
make # add '-j${N}' to specify number of threads to use for compilation
## wait for ITK to finish building
cmake -DCMAKE_INSTALL_PREFIX=${path_to_where_you_want_to_install} .. # this is for HistoReg
make # add '-j${N}' to specify number of threads to use for compilation
make install/strip # performs installation and executable stripping

Run

${installDir}/bin/HistoReg --help : Print help message
${installDir}/bin/HistoReg -m /path/to/moving/image -f /path/to/fix/image -o /path/to/output/dir/ -c /path/to/c2d/executable [-OPTIONAL]

Example

From the HistoReg directory, run :

${installDir}/bin/HistoReg -m ${HistoReg_source}/Data/Images/CD68.jpg -f ${HistoReg_source}/Data/Images/CD4.jpg -c /path/to/c2d/executable  -o ${HistoReg_source}/Data/Output/ -l ${HistoReg_source}/Data/Landmarks/CD68.csv -S

The code will perform affine and defformable registration from the moving (or source) image CD68.jpg (-m option) to the fixed ( or target) image CD4.jpg (-f option) and will store the result in the Output/ directory (-o option).

The -c option is to specify where the c2d executable is.

The -l and -S flags are optional, the first one is to apply the transformation on landmarks defined in the source image, and the second one is to apply the transformation on the resampled images to have an idea if the registration succeed or failed.

After the registration completed, you should find a folder called Output/CD68_registered_to_CD4 where all the result were saved :

• The output of the registration will be in the folder Output/CD68_registered_to_CD4/metrics where you will find the registration metrics computed at the resampled scale (small_resolution folder) and adapted to the original size of the images (full_resolution folder)
• The output of the -l command will be the file Output/CD68_registered_to_CD4/warped_landmarks.csv.
• The output of the -S command will be in the folder Output/CD68_registered_to_CD4/Saved_NIFTIs/small_resolution/, the source will be called new_small_source_padded.nii.gz, the target new_small_target_padded.nii.gz and the source registered to the target small_registeredImage.nii.gz, these images will be gray-scale images and padded after the preprocessing steps detailled in our paper (you can use the -F option to apply the transformation to the original RGB images at their original scale but the running time will be way longer)

To check if the registration worked, you can overlay these two images:

• new_small_target_padded.nii.gz
• small_registeredImage.nii.gz

The two images should be perfectly aligned. If you don't have a software to visualize these images, one viewer that can be used is ITK-SNAP, which was used for this project (download here).

You can then compare the warped landmarks (Output/CD68_registered_to_CD4/warped_landmarks.csv) with the landmarks defined in the target space (Data/Landmarks/CD4.csv), these landmarks should be very close (from a different of a few pixels to a few tenth of pixels, which is very small compared to the original images sizes).

Supported File types

• NIfTI
• PNG
• JPG
• TIFF

Note that all files are internally converted to NIfTI.

References

[1] L. Venet, S. Pati et al., "Accurate and Robust Alignment of Differently Stained Histologic Images Based on Greedy Diffeomorphic Registration," in Applied Sciences, 2021. DOI: 10.3390/app11041892.

[2] J. Borovec et al., "ANHIR: Automatic Non-rigid Histological Image Registration Challenge," in IEEE Transactions on Medical Imaging, 2020. DIO: 10.1109/TMI.2020.2986331.

[3] J. Borovec, A. Munoz-Barrutia, J. Kybic, "Benchmarking of Image Registration Methods for Differently Stained Histological Slides," 25th IEEE International Conference on Image Processing (ICIP), 2018. DOI: 10.1109/icip.2018.8451040