| About this Abstract |
| Meeting |
2024 TMS Annual Meeting & Exhibition
|
| Symposium
|
Novel Strategies for Rapid Acquisition and Processing of Large Datasets from Advanced Characterization Techniques
|
| Presentation Title |
Utilizing Deep Learning Techniques to Accelerate X-ray Absorption and Diffraction Contrast Imaging |
| Author(s) |
Eshan Ganju, Nikhilesh Chawla |
| On-Site Speaker (Planned) |
Nikhilesh Chawla |
| Abstract Scope |
In the past decade, laboratory-scale X-ray microscopy systems have enabled 3D tomographic and crystallographic imaging to solve a variety of different scientific problems. However, flux limitations in lab-scale systems often make time-resolved experiments challenging because the times required for X-ray exposure are very long. In this study, we present an in-depth analysis of a deep learning approach to expedite the acquisition of lab-scale absorption contrast tomography (ACT) and diffraction contrast tomography (DCT) data. We used low and high-dose ACT and DCT datasets to train conventional UNET++-based networks and more advanced Generative Adversarial Networks (GANs) with the goal of refining the low-dose data to approach the quality of the high-dose datasets. We also present a quantitative comparison of the image quality and time savings achieved via the use of DL-enhanced low-dose datasets. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Characterization, Machine Learning, Other |