Domain adaptation for segmentation of critical structures for prostate cancer therapy
Abstract Preoperative assessment of the proximity of critical structures to the tumors is crucial in avoiding unnecessary damage during prostate cancer treatment. A patient-specific 3D anatomical model of those structures, namely the neurovascular bundles (NVB) and the external urethral sphincters (...
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2021-06-01
|
| Series: | Scientific Reports |
| Online Access: | https://doi.org/10.1038/s41598-021-90294-4 |
| _version_ | 1818750748644605952 |
|---|---|
| author | Anneke Meyer Alireza Mehrtash Marko Rak Oleksii Bashkanov Bjoern Langbein Alireza Ziaei Adam S. Kibel Clare M. Tempany Christian Hansen Junichi Tokuda |
| author_facet | Anneke Meyer Alireza Mehrtash Marko Rak Oleksii Bashkanov Bjoern Langbein Alireza Ziaei Adam S. Kibel Clare M. Tempany Christian Hansen Junichi Tokuda |
| author_sort | Anneke Meyer |
| collection | DOAJ |
| description | Abstract Preoperative assessment of the proximity of critical structures to the tumors is crucial in avoiding unnecessary damage during prostate cancer treatment. A patient-specific 3D anatomical model of those structures, namely the neurovascular bundles (NVB) and the external urethral sphincters (EUS), can enable physicians to perform such assessments intuitively. As a crucial step to generate a patient-specific anatomical model from preoperative MRI in a clinical routine, we propose a multi-class automatic segmentation based on an anisotropic convolutional network. Our specific challenge is to train the network model on a unique source dataset only available at a single clinical site and deploy it to another target site without sharing the original images or labels. As network models trained on data from a single source suffer from quality loss due to the domain shift, we propose a semi-supervised domain adaptation (DA) method to refine the model’s performance in the target domain. Our DA method combines transfer learning and uncertainty guided self-learning based on deep ensembles. Experiments on the segmentation of the prostate, NVB, and EUS, show significant performance gain with the combination of those techniques compared to pure TL and the combination of TL with simple self-learning ( $${p}<0.005$$ p < 0.005 for all structures using a Wilcoxon’s signed-rank test). Results on a different task and data (Pancreas CT segmentation) demonstrate our method’s generic application capabilities. Our method has the advantage that it does not require any further data from the source domain, unlike the majority of recent domain adaptation strategies. This makes our method suitable for clinical applications, where the sharing of patient data is restricted. |
| first_indexed | 2024-12-18T04:24:36Z |
| format | Article |
| id | doaj.art-0b307645ba7c4525a5208fe3cfd85cda |
| institution | Directory Open Access Journal |
| issn | 2045-2322 |
| language | English |
| last_indexed | 2024-12-18T04:24:36Z |
| publishDate | 2021-06-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | Scientific Reports |
| spelling | doaj.art-0b307645ba7c4525a5208fe3cfd85cda2022-12-21T21:21:09ZengNature PortfolioScientific Reports2045-23222021-06-0111111410.1038/s41598-021-90294-4Domain adaptation for segmentation of critical structures for prostate cancer therapyAnneke Meyer0Alireza Mehrtash1Marko Rak2Oleksii Bashkanov3Bjoern Langbein4Alireza Ziaei5Adam S. Kibel6Clare M. Tempany7Christian Hansen8Junichi Tokuda9Department of Simulation and Graphics and Research Campus STIMULATE, University of MagdeburgDepartment of Radiology, Brigham and Women’s Hospital, Harvard Medical SchoolDepartment of Simulation and Graphics and Research Campus STIMULATE, University of MagdeburgDepartment of Simulation and Graphics and Research Campus STIMULATE, University of MagdeburgDepartment of Radiology, Brigham and Women’s Hospital, Harvard Medical SchoolDepartment of Radiology, Brigham and Women’s Hospital, Harvard Medical SchoolDivision of Urology, Department of Surgery, Brigham and Women’s Hospital, Harvard Medical SchoolDepartment of Radiology, Brigham and Women’s Hospital, Harvard Medical SchoolDepartment of Simulation and Graphics and Research Campus STIMULATE, University of MagdeburgDepartment of Radiology, Brigham and Women’s Hospital, Harvard Medical SchoolAbstract Preoperative assessment of the proximity of critical structures to the tumors is crucial in avoiding unnecessary damage during prostate cancer treatment. A patient-specific 3D anatomical model of those structures, namely the neurovascular bundles (NVB) and the external urethral sphincters (EUS), can enable physicians to perform such assessments intuitively. As a crucial step to generate a patient-specific anatomical model from preoperative MRI in a clinical routine, we propose a multi-class automatic segmentation based on an anisotropic convolutional network. Our specific challenge is to train the network model on a unique source dataset only available at a single clinical site and deploy it to another target site without sharing the original images or labels. As network models trained on data from a single source suffer from quality loss due to the domain shift, we propose a semi-supervised domain adaptation (DA) method to refine the model’s performance in the target domain. Our DA method combines transfer learning and uncertainty guided self-learning based on deep ensembles. Experiments on the segmentation of the prostate, NVB, and EUS, show significant performance gain with the combination of those techniques compared to pure TL and the combination of TL with simple self-learning ( $${p}<0.005$$ p < 0.005 for all structures using a Wilcoxon’s signed-rank test). Results on a different task and data (Pancreas CT segmentation) demonstrate our method’s generic application capabilities. Our method has the advantage that it does not require any further data from the source domain, unlike the majority of recent domain adaptation strategies. This makes our method suitable for clinical applications, where the sharing of patient data is restricted.https://doi.org/10.1038/s41598-021-90294-4 |
| spellingShingle | Anneke Meyer Alireza Mehrtash Marko Rak Oleksii Bashkanov Bjoern Langbein Alireza Ziaei Adam S. Kibel Clare M. Tempany Christian Hansen Junichi Tokuda Domain adaptation for segmentation of critical structures for prostate cancer therapy Scientific Reports |
| title | Domain adaptation for segmentation of critical structures for prostate cancer therapy |
| title_full | Domain adaptation for segmentation of critical structures for prostate cancer therapy |
| title_fullStr | Domain adaptation for segmentation of critical structures for prostate cancer therapy |
| title_full_unstemmed | Domain adaptation for segmentation of critical structures for prostate cancer therapy |
| title_short | Domain adaptation for segmentation of critical structures for prostate cancer therapy |
| title_sort | domain adaptation for segmentation of critical structures for prostate cancer therapy |
| url | https://doi.org/10.1038/s41598-021-90294-4 |
| work_keys_str_mv | AT annekemeyer domainadaptationforsegmentationofcriticalstructuresforprostatecancertherapy AT alirezamehrtash domainadaptationforsegmentationofcriticalstructuresforprostatecancertherapy AT markorak domainadaptationforsegmentationofcriticalstructuresforprostatecancertherapy AT oleksiibashkanov domainadaptationforsegmentationofcriticalstructuresforprostatecancertherapy AT bjoernlangbein domainadaptationforsegmentationofcriticalstructuresforprostatecancertherapy AT alirezaziaei domainadaptationforsegmentationofcriticalstructuresforprostatecancertherapy AT adamskibel domainadaptationforsegmentationofcriticalstructuresforprostatecancertherapy AT claremtempany domainadaptationforsegmentationofcriticalstructuresforprostatecancertherapy AT christianhansen domainadaptationforsegmentationofcriticalstructuresforprostatecancertherapy AT junichitokuda domainadaptationforsegmentationofcriticalstructuresforprostatecancertherapy |