X-ray computed tomography (CT) imaging technology, with nondestructive testing capabilities, has been widely used in industry, medicine, and other fields. When X-ray CT imaging is performed on samples containing high-absorption materials such as metals, the reconstructed images often contain metal artifacts due to beam hardening, scattering, and other factors, which severely degrade the quality of CT imaging. More recently, dual/multi-spectral CT has been proposed as an effective means of reducing beam-hardening and metal artifacts. However, it needs multiple scans of the object or specialized multi-spectral CT equipment. In this paper, we studied the multi-material decomposition reconstruction technique with traditional CT scanned data to reduce beam-hardening and metal artifacts.

The problem of multi-material decomposition reconstruction in traditional single-spectral CT is inherently highly underdetermined, leading to non-unique solutions. To obtain physically meaningful true solutions, it is necessary to incorporate additional constraints. In a type of scenario, the constituent materials of the scanned object are known and immiscible. The reconstructed image vectors are orthogonal if these materials are selected as basis materials needed in multi-material decomposition reconstruction. Based on this finding, an orthogonal multi-material decomposition reconstruction technique (OMDRT) combined with the X-ray energy spectrum was proposed. In the proposed OMDRT method, the order of basis materials was sorted based on the decreasing sequence of their attenuation coefficients. With triple-material decomposition reconstruction as an example, the proposed OMDRT method includes steps as follows: 1) triple-material decomposition reconstruction; 2) generation of the first material’ mark images from reconstructed image; 3) triple-material decomposition reconstruction with the first material’ mark images; 4) generation of the first and second materials’ mark images from reconstructed images; 5) triple-material decomposition reconstruction with the first and second materials’ mark images. Steps 4) and 5) were performed iteratively. In steps 3) and 5), the weights for the decomposition reconstruction of basis materials from the projection data were adjusted based on the materials’ regional location marked in the materials’ mark images。

The numerical phantom used in the simulation is shown in Fig. 2(c), and it includes three materials: water, bone (simulating the teeth), and AgHg (simulating the dental filling) with standard densities of 1 g/cm³, 1.92 g/cm³, and 12 g/cm³, respectively. If the mass attenuation coefficients of these three materials are used as basis functions, the density of the material region in the corresponding image is the standard density. We select AgHg as the first basis material, bone as the second basis material, and water as the third basis material. By using the simulated projections of phantom without and with noise, density images are reconstructed with the proposed OMDRT. From the last rows in Fig. 4 and Fig. 11, we can see that the three materials are mostly separated in the results of three iterations, and metal artifacts have been effectively corrected basically. Figure 8 and figure 13 show that there are no obvious artifacts in either the density images or the virtual monochromatic image. To quantitatively analyze the image quality, we calculate the peak signal-to-noise ratio (PSNR) and normalized mean absolute deviation (NMAD) between the resulting virtual monochromatic images and the actual virtual monochromatic images. From Fig. 7 and Fig. 12, we can observe that the proposed OMDRT method converges within several iterations. In summary, the experimental results show that the method proposed in this paper has a good application effect in reducing metal artifacts.

For the metal artifact correction in CT images of scanned objects with known and non-mixing materials, we propose an iterative OMDRT of traditional CT. The proposed method chooses known materials as the basis materials, adjusting the weights for the decomposition reconstruction of basis materials based on their regional location. We choose a dental phantom with dental fillings to verify the validity of the proposed method. The basis materials are separated correctly with our method for both simulated noise-free data and Poisson noise data. In addition, artifacts caused by metal implants in both the triple-basis density images and the virtual monochromatic images combined by them are reduced effectively. Moreover, the proposed method converges within a small number of iterations, facilitating its widespread practical application. We verify the multi-material decomposition reconstruction technique of traditional CT. The experimental part does not utilize actual data and does not consider the effect of scattered photons, which are issues that require further research. During the experimental process, it is found that the accuracy of the spectrum significantly affects the effectiveness of the proposed method. How to acquire spectrum quickly and accurately is also a challenge that needs to be addressed in practical experiments. Future work will cover the OMDRT of dual/multi-spectral CT and explore its effectiveness in other applications.