Physics-assisted generative adversarial network for X-ray tomography
<jats:p>X-ray tomography is capable of imaging the interior of objects in three dimensions non-invasively, with applications in biomedical imaging, materials science, electronic inspection, and other fields. The reconstruction process can be an ill-conditioned inverse problem, requiring regula...
| Main Authors: | , , , , , , , |
|---|---|
| Other Authors: | |
| Format: | Article |
| Language: | English |
| Published: |
Optica Publishing Group
2023
|
| Online Access: | https://hdl.handle.net/1721.1/150780 |
| _version_ | 1811078409180151808 |
|---|---|
| author | Guo, Zhen Song, Jung Ki Barbastathis, George Glinsky, Michael E Vaughan, Courtenay T Larson, Kurt W Alpert, Bradley K Levine, Zachary H |
| author2 | Massachusetts Institute of Technology. Department of Mechanical Engineering |
| author_facet | Massachusetts Institute of Technology. Department of Mechanical Engineering Guo, Zhen Song, Jung Ki Barbastathis, George Glinsky, Michael E Vaughan, Courtenay T Larson, Kurt W Alpert, Bradley K Levine, Zachary H |
| author_sort | Guo, Zhen |
| collection | MIT |
| description | <jats:p>X-ray tomography is capable of imaging the interior of objects in three dimensions non-invasively, with applications in biomedical imaging, materials science, electronic inspection, and other fields. The reconstruction process can be an ill-conditioned inverse problem, requiring regularization to obtain satisfactory results. Recently, deep learning has been adopted for tomographic reconstruction. Unlike iterative algorithms which require a distribution that is known <jats:italic>a priori</jats:italic>, deep reconstruction networks can learn a prior distribution through sampling the training distributions. In this work, we develop a Physics-assisted Generative Adversarial Network (PGAN), a two-step algorithm for tomographic reconstruction. In contrast to previous efforts, our PGAN utilizes maximum-likelihood estimates derived from the measurements to regularize the reconstruction with both known physics and the learned prior. Compared with methods with less physics assisting in training, PGAN can reduce the photon requirement with limited projection angles to achieve a given error rate. The advantages of using a physics-assisted learned prior in X-ray tomography may further enable low-photon nanoscale imaging.</jats:p> |
| first_indexed | 2024-09-23T10:59:02Z |
| format | Article |
| id | mit-1721.1/150780 |
| institution | Massachusetts Institute of Technology |
| language | English |
| last_indexed | 2024-09-23T10:59:02Z |
| publishDate | 2023 |
| publisher | Optica Publishing Group |
| record_format | dspace |
| spelling | mit-1721.1/1507802023-05-20T03:34:10Z Physics-assisted generative adversarial network for X-ray tomography Guo, Zhen Song, Jung Ki Barbastathis, George Glinsky, Michael E Vaughan, Courtenay T Larson, Kurt W Alpert, Bradley K Levine, Zachary H Massachusetts Institute of Technology. Department of Mechanical Engineering <jats:p>X-ray tomography is capable of imaging the interior of objects in three dimensions non-invasively, with applications in biomedical imaging, materials science, electronic inspection, and other fields. The reconstruction process can be an ill-conditioned inverse problem, requiring regularization to obtain satisfactory results. Recently, deep learning has been adopted for tomographic reconstruction. Unlike iterative algorithms which require a distribution that is known <jats:italic>a priori</jats:italic>, deep reconstruction networks can learn a prior distribution through sampling the training distributions. In this work, we develop a Physics-assisted Generative Adversarial Network (PGAN), a two-step algorithm for tomographic reconstruction. In contrast to previous efforts, our PGAN utilizes maximum-likelihood estimates derived from the measurements to regularize the reconstruction with both known physics and the learned prior. Compared with methods with less physics assisting in training, PGAN can reduce the photon requirement with limited projection angles to achieve a given error rate. The advantages of using a physics-assisted learned prior in X-ray tomography may further enable low-photon nanoscale imaging.</jats:p> 2023-05-19T13:52:37Z 2023-05-19T13:52:37Z 2022 2023-05-19T13:50:19Z Article http://purl.org/eprint/type/JournalArticle https://hdl.handle.net/1721.1/150780 Guo, Zhen, Song, Jung Ki, Barbastathis, George, Glinsky, Michael E, Vaughan, Courtenay T et al. 2022. "Physics-assisted generative adversarial network for X-ray tomography." Optics Express, 30 (13). en 10.1364/OE.460208 Optics Express Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use. application/pdf Optica Publishing Group Optica |
| spellingShingle | Guo, Zhen Song, Jung Ki Barbastathis, George Glinsky, Michael E Vaughan, Courtenay T Larson, Kurt W Alpert, Bradley K Levine, Zachary H Physics-assisted generative adversarial network for X-ray tomography |
| title | Physics-assisted generative adversarial network for X-ray tomography |
| title_full | Physics-assisted generative adversarial network for X-ray tomography |
| title_fullStr | Physics-assisted generative adversarial network for X-ray tomography |
| title_full_unstemmed | Physics-assisted generative adversarial network for X-ray tomography |
| title_short | Physics-assisted generative adversarial network for X-ray tomography |
| title_sort | physics assisted generative adversarial network for x ray tomography |
| url | https://hdl.handle.net/1721.1/150780 |
| work_keys_str_mv | AT guozhen physicsassistedgenerativeadversarialnetworkforxraytomography AT songjungki physicsassistedgenerativeadversarialnetworkforxraytomography AT barbastathisgeorge physicsassistedgenerativeadversarialnetworkforxraytomography AT glinskymichaele physicsassistedgenerativeadversarialnetworkforxraytomography AT vaughancourtenayt physicsassistedgenerativeadversarialnetworkforxraytomography AT larsonkurtw physicsassistedgenerativeadversarialnetworkforxraytomography AT alpertbradleyk physicsassistedgenerativeadversarialnetworkforxraytomography AT levinezacharyh physicsassistedgenerativeadversarialnetworkforxraytomography |