A preliminary study of characteristics of cross scattering for multi-source cone-beam CT

The development of multi-source cone-beam computed tomography (CBCT) systems has significantly improved imaging speed, making them crucial in various clinical and research applications, particularly in dynamic studies. However, the simultaneous exposure of multiple X-ray sources increases the complexity of scatter within the system. The detector receives not only forward scattering from the directly aligned source but also cross scattering from other sources. Traditional scatter correction methods are mainly designed for single-source or dual-source CT, which often fail to address the challenges posed by multi-source scenario.

The purpose of this work is to proposes a GPU-based Monte Carlo (MC) simulation method to accurately model and analyze the scatter distribution and the characteristics of different components in multi-source CBCT systems, as well as the impact of cross scattering on image reconstruction quality.

We utilize the GPU-based MC package gMCDRR to simulate multi-source CBCT system, including both flat-panel and spherical detector configurations. The simulation process consists of three main components: imaging system modeling, photon initialization, and the simulation of physical interactions in phantom. The whole process of photon interaction with the geometry and its arrival at the detector is simulated in parallel using multiple GPU cores to enhance computational efficiency. We compare the proposed method with Geant4 method to validate the simulation accuracy, the intensity of cross scattering and its impact on reconstructed images are also quantitively evaluated.

The simulation results show that the scattering intensity in multi-source CBCT increases with the number of X-ray sources. The cross scattering is mainly affected by the relative deflection angle between the sources and detector, as well as the characteristics of the imaging object. The backward scattering intensity at 180° potentially may exceed forward scattering, and this effect is more pronounced in larger phantoms. The reconstructed images from multi-source CBCT are severely affected by the superimposed cross scattering, resulting in increased artifacts and inaccurate CT values, which can tremendously degrade the image quality.

The proposed GPU-based MC simulation method can accurately model and analyze the scatter distribution and characteristics of different components in multi-source CBCT systems, which can be utilized to correct the cross scattering of the multi-source CBCT.