Elitomo Project: 2D &3D Tomography for Large-size Objects Using ELI-NP Ultrabright Gamma Beam Source

1. SHORT DESCRIPTION

Many of the important scientific and engineering achievements of the last decades can be traced back to non-destructive testing (NDT) methods that have been used during development stage without damaging or affecting the product properties, hence assuring the performance and integrity for which the product was intended. From all existent NDT methods, the classical radiography has been spectacularly improved lately thanks to the development of the microfocus X-ray sources and of very high-resolution X-ray detectors. As a result, two new improved sub-domains called Digital Radiography (DR) and Industrial Computed Tomography (ICT) have been developed and became the most advanced techniques that could offer high-penetrability at very high spatial resolution. Despite the latest improvements there is still a lack of high-resolution NDT solutions for investigating large size metallic components, such as aeronautics, space and military products, cars/trucks engine blocks and gearbox, oil refinery parts, etc. The industrial solutions available are based on using high-energy radiation sources in the range of 1-15 MeV (such as Linacs) or isotopic sources (Co60, for example). But these solutions deliver a very big focal spot size in the range of 0.5 – 2 mm [1], which reduces dramatically the spatial resolution and are practically not useful in the high-resolution applications. From all other techniques available for producing high-energy photons with small focal spot, the Laser Compton Scattering is of very much interest [2] that could meet all the requirements. This is the reason why we have proposed in the ELI White Book [3] to develop the digital radiography and tomography applications by using ELI Gamma beam, as a perfect solution for high-resolution DR and ICT investigation of large-size objects. It is noteworthy to mention the main advantages of using ELI Gamma Beam source in DR and ICT for industrial large-size objects for NDT applications. (i) It is the only solution available that could provide small focal spot size (10-30 microns) providing very high-resolution NDT imaging even for very large-size objects that are easily penetrated by the high-energy beam; (ii) By using photons over a few MeV the results are almost free from beam hardening effect and the metal streak artifacts usual found in classical bremsstrahlung DR and ICT techniques. (iii) By having a tunable monochromatic gamma beam we can measure the sample’s exact values of the linear attenuation coefficient and, by doing same measurement at different gamma energies, a dual/multi energy technique could be applied for material identification. At ELI-NP we expect to achieve unprecedented spatial resolution (100-200 microns) and contrast sensitivity in industrial applications since the intensity of the ELI-NP gamma beam will be few orders of magnitude higher than at HIγS (USA) or at AIST (Japan). Up to now, the ELITOMO partners have developed the Monte Carlo simulation for industrial NDT applications and established the basic requirements for the DR and ICT mechanical scanners. Within this project we propose to design and develop detection systems and to develop the dedicated scanning control procedure and the digital acquisition and read-out system for these DR and ICT setups before installation at ELI-NP. In addition, a series of preliminary tests and experiments will be made in laboratory by using X-Ray microfocus sources, gamma sources or LINAC. Finally a series of tests at CETAL or other LCS facilities will be followed by the installation of the two systems at ELI-NP GAMMA BEAM facility.

2. PROJECT OBJECTIVES

The ELITOMO project aims at providing support for industrial imaging applications development at ELI-NP gamma beam system. The project objectives are:

• Study the impact and the advantages of using X-ray sources (AP2K) and currently available gamma sources (Co-60, Subaru-Japan) as test beds for the experimental configurations using Monte Carlo simulations
• Develop dedicated scanning control procedure and digital acquisition and read-out system for pencil-beam and cone-beam tomography setups. Specific tomographic reconstruction algorithms will be developed and image analysis software packages will be tested.
• Detector development and prototyping for pencil beam experiments. The high intensity of the gamma beam and the peculiar gamma beam time structure yield high intensity photon pulses that arrive at the detector in a picosecond window. The high intensity photon pulses will create radiation damage in the conventional detectors and saturation of the PMTs. We propose to study the impact of using alternative detection schemes to the conventional NaI detectors coupled with PMT that are often used in the CT experiments. Using Monte Carlo simulations we propose to study the possibility to use a gamma-ray calorimeter for the pencil-beam experiments.
• Test both scanners and software packages in various laboratory environments with different type of sources: X-rays and currently available gamma ray sources before installation at ELI-NP
• Formation and training of experimental personnel for future operation of the experimental setups dedicated for tomography and radiography
• Offer support for installation and tests of both DR and ICT systems at ELI-NP

3. SUMMARY OF RESULTS

The performed activities up to now are:

• Implementation of LCS photon sources in Geant4 for assessing the response of the detectors in conditions like those in a real experiment (IHIN-HH);
• Development of two fan-beam tomographic reconstruction algorithms: Filtered Back Projection (FBP) and Algebraic Iterative Reconstruction algorithm (AIR) (AP2K)
• Monte Carlo simulations of line-pair structures projections for estimating the spatial resolution and contrast sensitivity for ELI-NP gamma beam environment (IHIN-HH and AP2K)
• Testing of the VOLUME GRAPHICS© software, a high-end software package for analysis and visualization of industrial computed tomography (CT) data (AP2K)
• Scintillators type analysis for optimization of the 2D detector for ELI-NP gamma-beam tomography (IHIN-HH)
• Setting up a cluster computer capable to perform parallel computing for Monte Carlo simulations, tomographic reconstruction and image analysis (AP2K)
• Using of VOLUME GRAPHICS© software package for performing 2D tomographic reconstruction (AP2K)
• Monte Carlo simulations with Geant4 for translation + rotation system for scanning larger dimension objects with ELINP Gamma Beam (AP2K)
• Performing contrast analysis, signal to noise (SNR) evaluation on Geant4 simulated data for targets of different materials and sizes (IFIN-HH and AP2K)
• 2D gamma detector development for gamma beam experiments (AP2K)
• Design and development of the scanning control procedure, data acquisition, data reconstruction algorithm integration for large objects in X-ray fan/pencil beam laboratory environment (AP2K)
• A gamma-ray calorimeter simulation and design for computed tomography gamma applications (IFIN-HH)

4. SCIENTIFIC RESULTS PRESENTATION

4.1 Simulations of LCS photon sources in Geant4 for assessing the response of the detectors in conditions like those in a real experiment (IHIN-HH)

The available LCS sources can provide gamma beams with several percent bandwidths (e.g. 3% at HIγS) and intensities of 105 photons/s at NewSubaru and 107 photons/s at HIγS. To assess the detector responses in conditions similar to the real experiments and to compare the performance of the DR and ICT setups with different sources we have implemented some of the above-mentioned gamma sources in our Geant4 simulation code. The monoenergetic sources and the X-ray/bremsstrahlung sources are implemented using the General Particle Source toolkit in Geant4, using the Gaussian and the bremsstrahlung distributions. An example of energy distributions for these two types of sources is displayed in figure 1.

4.2 Detector Response Simulations (IFIN-HH)

Geant4 simulations were also used to characterize the response of detectors to quasi-monoenergetic beams like the future gamma beam from ELI-NP.
Two types of detectors are foreseen to be used for gamma beam DR and ICT setups at ELI-NP: (i) a large volume detector capable of absorbing almost entirely the transmitted
gamma beam for pencil-beam CT, and (ii) a 2D detector for cone-beam CT consisting of a converter screen (converts gamma rays to visible photons) coupled via optic elements to a CCD camera. The deposited energies, the detector efficiencies, as well as the peak-to-
ratio were determined for different types and sizes of scintillators.
Figure 2 shows the peak to total ratio of a large volume scintillator and Table 2 lists the deposited energy in various thin scintillators

4.3 . Design and manufacturing of a cluster computer for parallel computing in Monte Carlo

For simulation and data reconstruction processes we have built a 96 cores computing cluster consisting of 4 units, each unit having 4 Intel Xeon X5650 processors and 16 GB RAM, each one running CENTOS 7 operating system coupled via MPI CH. In order take advantage of multiprocessing in GEANT4 we installed G4MPI (a native interface with MPI libraries simulation programs) and adjust the simulation program for parallel computing. This 96 cores architecture allowed to perform a Monte Carlo simulation of 1011 photons/projection and 100 projections, for an object having25 cm3, in about 4 days.

Fig. 3 96cores cluster computer used in AP2K simulations

4.4. Determination of Spatial Resolution and Contrast Sensitivity for Reconstructed Simulated Tomographic Data (AP2K)

For the tomographic data simulation, we used a fan beam set-up having 10m source-object distance, 40 m source-detector distance, 4x4 cm2 detector size, detector pitch of 0.1x0.1 mm2, the beam focal spot size of 10 µm and beam divergence of 1 mrad, all data simulated for 1 MeV gamma beam energy. The simulated objects were a 5 and 10 cm iron block having inside (a) iron cylinders containing cylindrical holes (air filled) of different sizes (fig. 4a), iron block containing spherical air holes (fig. 4b), respectively iron steps of different width containing cylindrical air holes (fig.4c).

Fig. 4 Simulated targets: A- iron cylinder with cylindrical holes, B-iron block with spherical holes, C- iron steps with cylindrical air holes.

Fig. 5 Metal block with spherical air holes. A - simulated spheres, B -simulated projection for 10cm Iron block with 10¹⁰ incident photons, C - simulated projection for 5 cm iron block with 10¹⁰ incident photons, D -simulated projection for 5 cm iron block with 10¹¹ incident photons.

Fig. 7 Variation of the signal to noise ratio with hole diameter (left) and the Modulation Transfer Function (left) for Geant4 simulation of iron steps with cylindrical air holes (Fig6 C).

4.5. Tomographic Reconstruction (AP2K)

The algorithms used in reconstruction were the Filtered Back Projection and the iterative SART algorithm for different number of projections. For Filtered back projection algorithm [8] the IRTm package [11] and Volume Graphics Studio Max. package [13,14,15] were used, while the AIR [12] toolkit was used for iterative reconstruction [9,10]. One of the targets was to determine the minimum number of projection and the detector integration time necessary for achieving at least 200 microns spatial resolution. We determined that, for 1 MeV gamma intensity of 2.5*107 incident photons/mrad, a reasonable image quality (where the 200 microns cylinders are clearly separated) could be obtained for a minimum of 1 second integration time for one projection and for a minimum of 100 equal-distributed projections acquired.

Fig. 8
Reconstructed images using 33, 66,100 projections with Filtered Back Projections and SART algorithms.

Fig. 9
Image analysis with Volume Graphics.

4.6. 2D gamma detector development for gamma beam experiments. (AP2K)

The high intensity photon pulses will create radiation damage and saturation in the conventional detectors. We built a 2D gamma detector by using a scintillator sheet, a mirror and a CCD camera mounted in a dark room frame presented in Figure

Fig.10
Two-dimensional gamma ray detector.

4.7. New iterative 3D reconstruction algorithm development (AP2K - under development)

For reducing the scanning time required by the FBP scanning algorithm (when a minimum 100 projections are necessary for having around 200 microns resolution) we are further investigating the results of applying iterative reconstruction algorithm with a reduced number of projection.

Fig.11
Computer application for tomographic data simulation, iterative reconstruction and visualizations for ELINP Gama Beam.For reducing the scanning time required by the FBP scanning algorithm (when a minimum 100 projections are necessary for having around 200 microns resolution) we are further investigating the results of applying iterative reconstruction algorithm with a reduced number of projection.

4.8. A gamma-ray calorimeter for computed tomography applications (IFIN-HH under development)

We are currently studying the option to use a gamma calorimeter for characterizing the transmitted beam in the CT experiments starting from a simplified version presented in reference [17]. The calorimeter prototype will use several layers of polyethylene (PE) – Si detector spaced by larger PE absorbers ( Fig. 12a). Each layer consists of 3 cm PE preceding a Si detector. Four larger PE blocks of 12.56 cm, 12.56 cm, 15.66 cm and 12.56 cm separate the PE-Si layers. Figure 3b shows the simulation result of deposited energy as a function of the energy of the gamma beam for a 5- and 22-layer device. The procedure to determine the beam energy and the beam intensity is similar to the procedure described in reference [18].

Fig.12.
(a) A diagram of a gamma calorimeter prototype to be used for CT experiments. (b) The longitudinal distribution of the energy deposited in the Si detectors for several energies of the gamma beam.