2019 Fiscal Year Annual Research Report
高速中性子線を用いた高性能ＣＴ再構築アルゴリズムの開発
| Project/Area Number |
18F18763
|
|---|
| Research Institution | Institute of Physical and Chemical Research |
|---|
Principal Investigator |
大竹 淑恵 国立研究開発法人理化学研究所, 光量子工学研究センター, チームリーダー (50216777)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
YAN MINGFEI 国立研究開発法人理化学研究所, 光量子工学研究センター, 外国人特別研究員
|
|---|
| Project Period (FY) |
2018-11-09 – 2021-03-31
|
| Keywords | Non-destructive test / Neutron / Linear scanning CT / System matrix / Image reconstruction |
|---|
| Outline of Annual Research Achievements |
To achieve the super resolution computer tomography (CT) with pixel-type fast neutron detector, the linear + rotation scan method is proposed. Point, line and area models are the three typical models for system matrix calculation. Theoretically, area model has the highest accuracy. In fast neutron CT experiment at RIKEN Accelerator-driven compact Neutron Source (RANS) with pixel-type detector (pixel: 2cm×2cm×5cm (T)), we hope to achieve ~mm space resolution. Because the pixel size is large and the number of pixels is small in the detector, the area model should be better than point and line models. Therefore, we established the area model for system matrix calculation. Firstly, we compared the results of system matrix with different models, and they show the summation of reshaped system matrix with area model is smoother than line model. Secondly, we compared the reconstructed results, and they also show area model is better than line model.
We tested the pixel-type fast neutron detector and implemented the CT experiment with acrylic bar on RANS. The preliminarily analyzed results shows the thickness of the bar (1.5cm) can be recognized by the linear scan method, which means the super resolution can be preliminary realized in the experiment. The detailed analysis is in progress.
We attended and made a poster presentation about rotation + linear scan CT from simulation reconstruction on the 8th international meeting of the Union for Compact Accelerator-driven Neutron Sources (UCANS-8) holding in Paris. We have also published a journal paper on EPJ Web of Conferences.
|
| Current Status of Research Progress |
Current Status of Research Progress
2: Research has progressed on the whole more than it was originally planned.
Reason
1. We have established the area model for system matrix calculation in rotation+linear scan fast neutron CT reconstruction. It is an extremely important part in achieving the super resolution CT with the pixel-type fast neutron detector. By means of the system matrix calculated with area model, we have successfully reconstructed 1cm diameter acrylic bar in the concrete with the area model and sparse reconstruction algorithm from the simulated projection data recorded by the fast neutron detector with 2cm×2cm×5cm (thickness) pixel size, which means the super resolution CT has been preliminary realized. In addition, we have made the comparison of reconstructed images with line model and area model. The results show that area model is better in rotation+linear scan fast neutron CT reconstruction.
2. At present, we have finished the rotation+linear scan fast neutron CT experiment on compact neutron source with pixel-type fast neutron detector. From the preliminarily analyzed results, the thickness of the bar (1.5 cm) can be recognized by the linear scan method with pixel-type detector (pixel size 2cm×2cm×5cm (thickness)), which means the super resolution can be preliminarily realized in the experiment. After the analysis of experimental data is finished, we can publish a paper on the topic of rotation+linear scan fast neutron CT, which mainly consists of projection data simulation by Geant4, system matrix calculation with area model, image reconstruction and experimental verification.
|
| Strategy for Future Research Activity |
1. At present, we have realized the system matrix with area model. But the deeply understanding of it when applied in rotation+linear scan fast neutron CT still needs to be continued. We plan to study the feature of the system matrix as the scanning parameters change (such as, the translation step, translation times, rotation step, rotation times and so on). In addition, the quantitatively evaluation method of the accuracy of the reconstructed image should be determined. We plan to use the some parameters, such as,
correlation coefficient, normalized root mean error, and so on, to evaluate it, where the pixel value of the reference image could be regarded as the linear attenuation coefficient of neutron attenuation in the matter.
2. At present, when we analyze the experimental data, the main challenge is to find the method to solve the problem of inconsistent response of different pixels of the fast neutron detector. We plan to make a correction coefficient for different pixels by analyzing and summarizing the experiment data, as well as referring to the simulation data. In making the correction coefficient, both of the neutron and gamma-ray factors will be considered.
Once the correction coefficient is found, we can finish the rotation+linear scan fast neutron CT data analysis.
|
-
[Journal Article] Reconstruction on fast neutron CT for concrete structure inspection with a pixel-type detector by applying linear scanning method2020
Author(s)
Mingfei Yan, Yasuo Wakabayashi, Yoshie Otake, Yujiro Ikeda, Atsushi Taketani, Takao Hashiguchi, Sheng Wang, Binbin Tian, Takaoki Takanashi, Tomohiro Kobayashi, and Baolong Ma
-
Journal Title
EPJ Web of Conferences
Volume: 231
Pages: 05008-1 -4v
DOI
Peer Reviewed / Open Access / Int'l Joint Research
-
