Scanner invariant representations for diffusion MRI harmonization. Moyer, D., Ver Steeg, G., Tax, C. M. W., & Thompson, P. M. Magnetic Resonance in Medicine, 84(4):2174-2189, 2020. ![Scanner invariant representations for diffusion MRI harmonization [link]](https://bibbase.org/img/filetypes/link.svg "link")
Paper doi abstract bibtex Purpose In the present work, we describe the correction of diffusion-weighted MRI for site and scanner biases using a novel method based on invariant representation. Theory and Methods Pooled imaging data from multiple sources are subject to variation between the sources. Correcting for these biases has become very important as imaging studies increase in size and multi-site cases become more common. We propose learning an intermediate representation invariant to site/protocol variables, a technique adapted from information theory-based algorithmic fairness; by leveraging the data processing inequality, such a representation can then be used to create an image reconstruction that is uninformative of its original source, yet still faithful to underlying structures. To implement this, we use a deep learning method based on variational auto-encoders (VAE) to construct scanner invariant encodings of the imaging data. Results To evaluate our method, we use training data from the 2018 MICCAI Computational Diffusion MRI (CDMRI) Challenge Harmonization dataset. Our proposed method shows improvements on independent test data relative to a recently published baseline method on each subtask, mapping data from three different scanning contexts to and from one separate target scanning context. Conclusions As imaging studies continue to grow, the use of pooled multi-site imaging will similarly increase. Invariant representation presents a strong candidate for the harmonization of these data.
@article{dan_mrm,
Abstract = {Purpose In the present work, we describe the correction of diffusion-weighted MRI for site and scanner biases using a novel method based on invariant representation. Theory and Methods Pooled imaging data from multiple sources are subject to variation between the sources. Correcting for these biases has become very important as imaging studies increase in size and multi-site cases become more common. We propose learning an intermediate representation invariant to site/protocol variables, a technique adapted from information theory-based algorithmic fairness; by leveraging the data processing inequality, such a representation can then be used to create an image reconstruction that is uninformative of its original source, yet still faithful to underlying structures. To implement this, we use a deep learning method based on variational auto-encoders (VAE) to construct scanner invariant encodings of the imaging data. Results To evaluate our method, we use training data from the 2018 MICCAI Computational Diffusion MRI (CDMRI) Challenge Harmonization dataset. Our proposed method shows improvements on independent test data relative to a recently published baseline method on each subtask, mapping data from three different scanning contexts to and from one separate target scanning context. Conclusions As imaging studies continue to grow, the use of pooled multi-site imaging will similarly increase. Invariant representation presents a strong candidate for the harmonization of these data.},
Author = {Moyer, Daniel and {Ver Steeg}, Greg and Tax, Chantal M. W. and Thompson, Paul M.},
Date-Added = {2020-07-16 15:04:21 -0700},
Date-Modified = {2020-07-16 15:05:00 -0700},
Doi = {10.1002/mrm.28243},
Eprint = {https://onlinelibrary.wiley.com/doi/pdf/10.1002/mrm.28243},
Journal = {Magnetic Resonance in Medicine},
Keywords = {diffusion MRI, harmonization, invariant representation},
Number = {4},
Pages = {2174-2189},
Title = {Scanner invariant representations for diffusion MRI harmonization},
Url = {https://onlinelibrary.wiley.com/doi/abs/10.1002/mrm.28243},
Volume = {84},
Year = {2020},
Bdsk-Url-1 = {https://onlinelibrary.wiley.com/doi/abs/10.1002/mrm.28243},
Bdsk-Url-2 = {https://doi.org/10.1002/mrm.28243}}
Downloads: 0
{"_id":"RkAgsnS3TFTew9wwd","bibbaseid":"moyer-versteeg-tax-thompson-scannerinvariantrepresentationsfordiffusionmriharmonization-2020","author_short":["Moyer, D.","Ver Steeg, G.","Tax, C. M. W.","Thompson, P. M."],"bibdata":{"bibtype":"article","type":"article","abstract":"Purpose In the present work, we describe the correction of diffusion-weighted MRI for site and scanner biases using a novel method based on invariant representation. Theory and Methods Pooled imaging data from multiple sources are subject to variation between the sources. Correcting for these biases has become very important as imaging studies increase in size and multi-site cases become more common. We propose learning an intermediate representation invariant to site/protocol variables, a technique adapted from information theory-based algorithmic fairness; by leveraging the data processing inequality, such a representation can then be used to create an image reconstruction that is uninformative of its original source, yet still faithful to underlying structures. To implement this, we use a deep learning method based on variational auto-encoders (VAE) to construct scanner invariant encodings of the imaging data. Results To evaluate our method, we use training data from the 2018 MICCAI Computational Diffusion MRI (CDMRI) Challenge Harmonization dataset. Our proposed method shows improvements on independent test data relative to a recently published baseline method on each subtask, mapping data from three different scanning contexts to and from one separate target scanning context. Conclusions As imaging studies continue to grow, the use of pooled multi-site imaging will similarly increase. Invariant representation presents a strong candidate for the harmonization of these data.","author":[{"propositions":[],"lastnames":["Moyer"],"firstnames":["Daniel"],"suffixes":[]},{"propositions":[],"lastnames":["Ver Steeg"],"firstnames":["Greg"],"suffixes":[]},{"propositions":[],"lastnames":["Tax"],"firstnames":["Chantal","M.","W."],"suffixes":[]},{"propositions":[],"lastnames":["Thompson"],"firstnames":["Paul","M."],"suffixes":[]}],"date-added":"2020-07-16 15:04:21 -0700","date-modified":"2020-07-16 15:05:00 -0700","doi":"10.1002/mrm.28243","eprint":"https://onlinelibrary.wiley.com/doi/pdf/10.1002/mrm.28243","journal":"Magnetic Resonance in Medicine","keywords":"diffusion MRI, harmonization, invariant representation","number":"4","pages":"2174-2189","title":"Scanner invariant representations for diffusion MRI harmonization","url":"https://onlinelibrary.wiley.com/doi/abs/10.1002/mrm.28243","volume":"84","year":"2020","bdsk-url-1":"https://onlinelibrary.wiley.com/doi/abs/10.1002/mrm.28243","bdsk-url-2":"https://doi.org/10.1002/mrm.28243","bibtex":"@article{dan_mrm,\n\tAbstract = {Purpose In the present work, we describe the correction of diffusion-weighted MRI for site and scanner biases using a novel method based on invariant representation. Theory and Methods Pooled imaging data from multiple sources are subject to variation between the sources. Correcting for these biases has become very important as imaging studies increase in size and multi-site cases become more common. We propose learning an intermediate representation invariant to site/protocol variables, a technique adapted from information theory-based algorithmic fairness; by leveraging the data processing inequality, such a representation can then be used to create an image reconstruction that is uninformative of its original source, yet still faithful to underlying structures. To implement this, we use a deep learning method based on variational auto-encoders (VAE) to construct scanner invariant encodings of the imaging data. Results To evaluate our method, we use training data from the 2018 MICCAI Computational Diffusion MRI (CDMRI) Challenge Harmonization dataset. Our proposed method shows improvements on independent test data relative to a recently published baseline method on each subtask, mapping data from three different scanning contexts to and from one separate target scanning context. Conclusions As imaging studies continue to grow, the use of pooled multi-site imaging will similarly increase. Invariant representation presents a strong candidate for the harmonization of these data.},\n\tAuthor = {Moyer, Daniel and {Ver Steeg}, Greg and Tax, Chantal M. W. and Thompson, Paul M.},\n\tDate-Added = {2020-07-16 15:04:21 -0700},\n\tDate-Modified = {2020-07-16 15:05:00 -0700},\n\tDoi = {10.1002/mrm.28243},\n\tEprint = {https://onlinelibrary.wiley.com/doi/pdf/10.1002/mrm.28243},\n\tJournal = {Magnetic Resonance in Medicine},\n\tKeywords = {diffusion MRI, harmonization, invariant representation},\n\tNumber = {4},\n\tPages = {2174-2189},\n\tTitle = {Scanner invariant representations for diffusion MRI harmonization},\n\tUrl = {https://onlinelibrary.wiley.com/doi/abs/10.1002/mrm.28243},\n\tVolume = {84},\n\tYear = {2020},\n\tBdsk-Url-1 = {https://onlinelibrary.wiley.com/doi/abs/10.1002/mrm.28243},\n\tBdsk-Url-2 = {https://doi.org/10.1002/mrm.28243}}\n\n","author_short":["Moyer, D.","Ver Steeg, G.","Tax, C. M. W.","Thompson, P. M."],"bibbaseid":"moyer-versteeg-tax-thompson-scannerinvariantrepresentationsfordiffusionmriharmonization-2020","role":"author","urls":{"Paper":"https://onlinelibrary.wiley.com/doi/abs/10.1002/mrm.28243"},"keyword":["diffusion MRI","harmonization","invariant representation"],"metadata":{"authorlinks":{}}},"bibtype":"article","biburl":"https://bibbase.org/f/xkceWam7dm8sQkdAg/gregv.bib","dataSources":["AurbPYNYbHmtrsQ7q","WKAs5u54exPpcdrJw"],"keywords":["diffusion mri","harmonization","invariant representation"],"search_terms":["scanner","invariant","representations","diffusion","mri","harmonization","moyer","ver steeg","tax","thompson"],"title":"Scanner invariant representations for diffusion MRI harmonization","year":2020}