A domain adaptation benchmark for T1-weighted brain magnetic resonance image segmentation
Accurate brain segmentation is critical for magnetic resonance imaging (MRI) analysis pipelines. Machine-learning-based brain MR image segmentation methods are among the state-of-the-art techniques for this task. Nevertheless, the segmentations produced by machine learning models often degrade in th...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2022-09-01
|
| Series: | Frontiers in Neuroinformatics |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/fninf.2022.919779/full |
| _version_ | 1811266047017222144 |
|---|---|
| author | Parisa Saat Nikita Nogovitsyn Nikita Nogovitsyn Muhammad Yusuf Hassan Muhammad Yusuf Hassan Muhammad Athar Ganaie Muhammad Athar Ganaie Roberto Souza Roberto Souza Hadi Hemmati Hadi Hemmati |
| author_facet | Parisa Saat Nikita Nogovitsyn Nikita Nogovitsyn Muhammad Yusuf Hassan Muhammad Yusuf Hassan Muhammad Athar Ganaie Muhammad Athar Ganaie Roberto Souza Roberto Souza Hadi Hemmati Hadi Hemmati |
| author_sort | Parisa Saat |
| collection | DOAJ |
| description | Accurate brain segmentation is critical for magnetic resonance imaging (MRI) analysis pipelines. Machine-learning-based brain MR image segmentation methods are among the state-of-the-art techniques for this task. Nevertheless, the segmentations produced by machine learning models often degrade in the presence of expected domain shifts between the test and train sets data distributions. These domain shifts are expected due to several factors, such as scanner hardware and software differences, technology updates, and differences in MRI acquisition parameters. Domain adaptation (DA) methods can make machine learning models more resilient to these domain shifts. This paper proposes a benchmark for investigating DA techniques for brain MR image segmentation using data collected across sites with scanners from different vendors (Philips, Siemens, and General Electric). Our work provides labeled data, publicly available source code for a set of baseline and DA models, and a benchmark for assessing different brain MR image segmentation techniques. We applied the proposed benchmark to evaluate two segmentation tasks: skull-stripping; and white-matter, gray-matter, and cerebrospinal fluid segmentation, but the benchmark can be extended to other brain structures. Our main findings during the development of this benchmark are that there is not a single DA technique that consistently outperforms others, and hyperparameter tuning and computational times for these methods still pose a challenge before broader adoption of these methods in the clinical practice. |
| first_indexed | 2024-04-12T20:34:57Z |
| format | Article |
| id | doaj.art-808b6bef5fe04c77bc6a4a50591c1029 |
| institution | Directory Open Access Journal |
| issn | 1662-5196 |
| language | English |
| last_indexed | 2024-04-12T20:34:57Z |
| publishDate | 2022-09-01 |
| publisher | Frontiers Media S.A. |
| record_format | Article |
| series | Frontiers in Neuroinformatics |
| spelling | doaj.art-808b6bef5fe04c77bc6a4a50591c10292022-12-22T03:17:36ZengFrontiers Media S.A.Frontiers in Neuroinformatics1662-51962022-09-011610.3389/fninf.2022.919779919779A domain adaptation benchmark for T1-weighted brain magnetic resonance image segmentationParisa Saat0Nikita Nogovitsyn1Nikita Nogovitsyn2Muhammad Yusuf Hassan3Muhammad Yusuf Hassan4Muhammad Athar Ganaie5Muhammad Athar Ganaie6Roberto Souza7Roberto Souza8Hadi Hemmati9Hadi Hemmati10Electrical and Software Engineering, Schulich School of Engineering, University of Calgary, Calgary, AB, CanadaCentre for Depression and Suicide Studies, St. Michael's Hospital, Toronto, ON, CanadaMood Disorders Program, Department of Psychiatry and Behavioural Neurosciences, McMaster University, Hamilton, ON, CanadaElectrical and Software Engineering, Schulich School of Engineering, University of Calgary, Calgary, AB, CanadaElectrical Engineering, Indian Institute of Technology, Gandhinagar, Gujarat, IndiaElectrical and Software Engineering, Schulich School of Engineering, University of Calgary, Calgary, AB, CanadaChemical Engineering, Indian Institute of Technology, Kharagpur, West Bengal, IndiaElectrical and Software Engineering, Schulich School of Engineering, University of Calgary, Calgary, AB, CanadaHotchkiss Brain Institute, Cumming School of Medicine, University of Calgary, Calgary, AB, CanadaElectrical and Software Engineering, Schulich School of Engineering, University of Calgary, Calgary, AB, CanadaElectrical Engineering and Computer Science, Lassonde School of Engineering, York University, Toronto, ON, CanadaAccurate brain segmentation is critical for magnetic resonance imaging (MRI) analysis pipelines. Machine-learning-based brain MR image segmentation methods are among the state-of-the-art techniques for this task. Nevertheless, the segmentations produced by machine learning models often degrade in the presence of expected domain shifts between the test and train sets data distributions. These domain shifts are expected due to several factors, such as scanner hardware and software differences, technology updates, and differences in MRI acquisition parameters. Domain adaptation (DA) methods can make machine learning models more resilient to these domain shifts. This paper proposes a benchmark for investigating DA techniques for brain MR image segmentation using data collected across sites with scanners from different vendors (Philips, Siemens, and General Electric). Our work provides labeled data, publicly available source code for a set of baseline and DA models, and a benchmark for assessing different brain MR image segmentation techniques. We applied the proposed benchmark to evaluate two segmentation tasks: skull-stripping; and white-matter, gray-matter, and cerebrospinal fluid segmentation, but the benchmark can be extended to other brain structures. Our main findings during the development of this benchmark are that there is not a single DA technique that consistently outperforms others, and hyperparameter tuning and computational times for these methods still pose a challenge before broader adoption of these methods in the clinical practice.https://www.frontiersin.org/articles/10.3389/fninf.2022.919779/fulldeep learningdomain adaptationmagnetic resonance imagingneuroimagingsegmentationbrain |
| spellingShingle | Parisa Saat Nikita Nogovitsyn Nikita Nogovitsyn Muhammad Yusuf Hassan Muhammad Yusuf Hassan Muhammad Athar Ganaie Muhammad Athar Ganaie Roberto Souza Roberto Souza Hadi Hemmati Hadi Hemmati A domain adaptation benchmark for T1-weighted brain magnetic resonance image segmentation Frontiers in Neuroinformatics deep learning domain adaptation magnetic resonance imaging neuroimaging segmentation brain |
| title | A domain adaptation benchmark for T1-weighted brain magnetic resonance image segmentation |
| title_full | A domain adaptation benchmark for T1-weighted brain magnetic resonance image segmentation |
| title_fullStr | A domain adaptation benchmark for T1-weighted brain magnetic resonance image segmentation |
| title_full_unstemmed | A domain adaptation benchmark for T1-weighted brain magnetic resonance image segmentation |
| title_short | A domain adaptation benchmark for T1-weighted brain magnetic resonance image segmentation |
| title_sort | domain adaptation benchmark for t1 weighted brain magnetic resonance image segmentation |
| topic | deep learning domain adaptation magnetic resonance imaging neuroimaging segmentation brain |
| url | https://www.frontiersin.org/articles/10.3389/fninf.2022.919779/full |
| work_keys_str_mv | AT parisasaat adomainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation AT nikitanogovitsyn adomainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation AT nikitanogovitsyn adomainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation AT muhammadyusufhassan adomainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation AT muhammadyusufhassan adomainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation AT muhammadatharganaie adomainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation AT muhammadatharganaie adomainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation AT robertosouza adomainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation AT robertosouza adomainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation AT hadihemmati adomainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation AT hadihemmati adomainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation AT parisasaat domainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation AT nikitanogovitsyn domainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation AT nikitanogovitsyn domainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation AT muhammadyusufhassan domainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation AT muhammadyusufhassan domainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation AT muhammadatharganaie domainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation AT muhammadatharganaie domainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation AT robertosouza domainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation AT robertosouza domainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation AT hadihemmati domainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation AT hadihemmati domainadaptationbenchmarkfort1weightedbrainmagneticresonanceimagesegmentation |