TY - JOUR
T1 - Ultrasparse View X-ray Computed Tomography for 4D Imaging
AU - Zheng, Yanjie
AU - Hatzell, Kelsey B.
N1 - Funding Information:
This material is based upon research supported by the Department of Energy, Solar Energy Technology Office under Award Number DE-EE0009385. K.B.H. and Y.Z. acknowledge support from the National Science Foundation under grant number 2140472.
Publisher Copyright:
© 2023 American Chemical Society.
PY - 2023/7/26
Y1 - 2023/7/26
N2 - X-ray computed tomography (CT) is a noninvasive, nondestructive approach to imaging materials, material systems, and engineered components in two and three dimensions. Acquisition of three-dimensional (3D) images requires the collection of hundreds or thousands of through-thickness X-ray radiographic images from different angles. Such 3D data acquisition strategies commonly involve suboptimal temporal sampling for in situ and operando studies (4D imaging). Herein, we introduce a sparse-view imaging approach, Tomo-NeRF, which is capable of reconstructing high-fidelity 3D images from <10 two-dimensional radiographic images. Experimental 2D and 3D X-ray images were used to test the reconstruction capability in two-view, four-view, and six-view scenarios. Tomo-NeRF is capable of reconstructing 3D images with a structural similarity of 0.9971-0.9975 and a voxel-wise accuracy of 81.83-89.59% from 2D experimentally obtained images. The reconstruction accuracy for the experimentally obtained images is less than the synthetic structures. Experimentally obtained images demonstrate a similarity of 0.9973-0.9984 and a voxel-wise accuracy of 84.31-95.77%.
AB - X-ray computed tomography (CT) is a noninvasive, nondestructive approach to imaging materials, material systems, and engineered components in two and three dimensions. Acquisition of three-dimensional (3D) images requires the collection of hundreds or thousands of through-thickness X-ray radiographic images from different angles. Such 3D data acquisition strategies commonly involve suboptimal temporal sampling for in situ and operando studies (4D imaging). Herein, we introduce a sparse-view imaging approach, Tomo-NeRF, which is capable of reconstructing high-fidelity 3D images from <10 two-dimensional radiographic images. Experimental 2D and 3D X-ray images were used to test the reconstruction capability in two-view, four-view, and six-view scenarios. Tomo-NeRF is capable of reconstructing 3D images with a structural similarity of 0.9971-0.9975 and a voxel-wise accuracy of 81.83-89.59% from 2D experimentally obtained images. The reconstruction accuracy for the experimentally obtained images is less than the synthetic structures. Experimentally obtained images demonstrate a similarity of 0.9973-0.9984 and a voxel-wise accuracy of 84.31-95.77%.
KW - X-ray computed tomography
KW - neural radiance field
KW - operando imaging
KW - sparse reconstruction
KW - tomographic reconstruction
UR - http://www.scopus.com/inward/record.url?scp=85165867230&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85165867230&partnerID=8YFLogxK
U2 - 10.1021/acsami.3c06291
DO - 10.1021/acsami.3c06291
M3 - Article
C2 - 37438943
AN - SCOPUS:85165867230
SN - 1944-8244
VL - 15
SP - 35024
EP - 35033
JO - ACS Applied Materials and Interfaces
JF - ACS Applied Materials and Interfaces
IS - 29
ER -