A comprehensive scatter correction model for dual-source computed tomography: Combining ambient scatter, cross-scatter, and forward scatter compensation

Compared to single-source computed tomography (CT), dual-source CT equipped with two cross-distributed X-ray beams enhances temporal resolution and acquires more comprehensive volumetric data. Nevertheless, the interaction between the two X-ray beams introduces more complex scatter signals into the acquired projection data. Existing methods typically model these scatter signals as the sum of cross-scatter and forward scatter, with cross-scatter estimation limited to single scatter along primary paths. We performed measurements on both phantoms and real objects using a custom-built dual-source CT at 120 kV. Results from a Yin-Yang phantom (9.6 cm in diameter) revealed that the peak ratio of hardware-induced ambient scatter to single-source projection intensity exceeds 60%, a factor often overlooked in existing methods. To achieve high-precision imaging in dual-source CT, it is essential to account for the ambient scatter component in acquired projection intensity. Therefore, we develop a more comprehensive model that decomposes the total scatter signals into three distinct components: ambient scatter, cross-scatter, and forward scatter. During the model solution process, we propose a cross-scatter kernel superposition (xSKS) module to enhance the accuracy of cross-scatter estimation by additionally modeling cross-scatter events along non-primary paths. Meanwhile, we introduce a fast object-adaptive scatter kernel superposition (FOSKS) module for efficient forward scatter estimation. In Monte Carlo (MC) simulations conducted on a modified numerical Catphan® 500 phantom, our model achieves a scatter-to-primary-weighted mean absolute percentage error (SPMAPE) of 1.32%, which is significantly lower than the 12.99% achieved by the state-of-the-art method. Physical experiments further validate its superior scatter artifact correction capability.