Cross Data Set Generalization of Ultrasound Image Augmentation using Representation Learning: A Case Study
Data augmentation is a common method to make deep learning assessible on limited data sets. However, classical image augmentation methods result in highly unrealistic images on ultrasound data. Another approach is to utilize learning-based augmentation methods, e.g. based on variational autoencoders...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
De Gruyter
2021-10-01
|
| Series: | Current Directions in Biomedical Engineering |
| Subjects: | |
| Online Access: | https://doi.org/10.1515/cdbme-2021-2193 |
| _version_ | 1797989317618958336 |
|---|---|
| author | Wulff Daniel Mehdi Mohamad Ernst Floris Hagenah Jannis |
| author_facet | Wulff Daniel Mehdi Mohamad Ernst Floris Hagenah Jannis |
| author_sort | Wulff Daniel |
| collection | DOAJ |
| description | Data augmentation is a common method to make deep learning assessible on limited data sets. However, classical image augmentation methods result in highly unrealistic images on ultrasound data. Another approach is to utilize learning-based augmentation methods, e.g. based on variational autoencoders or generative adversarial networks. However, a large amount of data is necessary to train these models, which is typically not available in scenarios where data augmentation is needed. One solution for this problem could be a transfer of augmentation models between different medical imaging data sets. In this work, we present a qualitative study of the cross data set generalization performance of different learning-based augmentation methods for ultrasound image data. We could show that knowledge transfer is possible in ultrasound image augmentation and that the augmentation partially results in semantically meaningful transfers of structures, e.g. vessels, across domains. |
| first_indexed | 2024-04-11T08:18:25Z |
| format | Article |
| id | doaj.art-a8dbfd825a6f4b0cb7e1018a3f68d09c |
| institution | Directory Open Access Journal |
| issn | 2364-5504 |
| language | English |
| last_indexed | 2024-04-11T08:18:25Z |
| publishDate | 2021-10-01 |
| publisher | De Gruyter |
| record_format | Article |
| series | Current Directions in Biomedical Engineering |
| spelling | doaj.art-a8dbfd825a6f4b0cb7e1018a3f68d09c2022-12-22T04:35:04ZengDe GruyterCurrent Directions in Biomedical Engineering2364-55042021-10-017275575810.1515/cdbme-2021-2193Cross Data Set Generalization of Ultrasound Image Augmentation using Representation Learning: A Case StudyWulff Daniel0Mehdi Mohamad1Ernst Floris2Hagenah Jannis3Institute for Robotics and Cognitive Systems, University of Luebeck, Ratzeburger Allee 160, 23562Luebeck, GermanyInstitute for Robotics and Cognitive Systems, University of Luebeck, Ratzeburger Allee 160, 23562Luebeck, GermanyInstitute for Robotics and Cognitive Systems, University of Luebeck, Ratzeburger Allee 160, 23562Luebeck, GermanyComputational Health Informatics Lab, Institute of Biomedical Engineering, University of Oxford, Old Road Campus Research Building,OxfordOX3 7DQ, UKData augmentation is a common method to make deep learning assessible on limited data sets. However, classical image augmentation methods result in highly unrealistic images on ultrasound data. Another approach is to utilize learning-based augmentation methods, e.g. based on variational autoencoders or generative adversarial networks. However, a large amount of data is necessary to train these models, which is typically not available in scenarios where data augmentation is needed. One solution for this problem could be a transfer of augmentation models between different medical imaging data sets. In this work, we present a qualitative study of the cross data set generalization performance of different learning-based augmentation methods for ultrasound image data. We could show that knowledge transfer is possible in ultrasound image augmentation and that the augmentation partially results in semantically meaningful transfers of structures, e.g. vessels, across domains.https://doi.org/10.1515/cdbme-2021-21932d ultrasoundvariational autoencodergenerative adversarial networklatent space |
| spellingShingle | Wulff Daniel Mehdi Mohamad Ernst Floris Hagenah Jannis Cross Data Set Generalization of Ultrasound Image Augmentation using Representation Learning: A Case Study Current Directions in Biomedical Engineering 2d ultrasound variational autoencoder generative adversarial network latent space |
| title | Cross Data Set Generalization of Ultrasound Image Augmentation using Representation Learning: A Case Study |
| title_full | Cross Data Set Generalization of Ultrasound Image Augmentation using Representation Learning: A Case Study |
| title_fullStr | Cross Data Set Generalization of Ultrasound Image Augmentation using Representation Learning: A Case Study |
| title_full_unstemmed | Cross Data Set Generalization of Ultrasound Image Augmentation using Representation Learning: A Case Study |
| title_short | Cross Data Set Generalization of Ultrasound Image Augmentation using Representation Learning: A Case Study |
| title_sort | cross data set generalization of ultrasound image augmentation using representation learning a case study |
| topic | 2d ultrasound variational autoencoder generative adversarial network latent space |
| url | https://doi.org/10.1515/cdbme-2021-2193 |
| work_keys_str_mv | AT wulffdaniel crossdatasetgeneralizationofultrasoundimageaugmentationusingrepresentationlearningacasestudy AT mehdimohamad crossdatasetgeneralizationofultrasoundimageaugmentationusingrepresentationlearningacasestudy AT ernstfloris crossdatasetgeneralizationofultrasoundimageaugmentationusingrepresentationlearningacasestudy AT hagenahjannis crossdatasetgeneralizationofultrasoundimageaugmentationusingrepresentationlearningacasestudy |