Local models for scatter estimation and descattering in polyenergetic X-ray tomography

Area of Science:

• Medical Imaging
• Computational Physics
• Image Reconstruction

Background:

• X-ray computed tomography (CT) is crucial in medical and industrial fields.
• X-ray scatter degrades CT image quality, causing contrast loss and artifacts.
• Accurate quantitative reconstructions are often hindered by uncorrected scatter.

Purpose of the Study:

• To introduce a novel modeling approach for scatter estimation and descattering in polyenergetic X-ray CT.
• To improve the quantitative accuracy of CT reconstructions by mitigating scatter effects.

Main Methods:

• Generated a training dataset of 2D radiographs with and without scatter using particle transport simulations.
• Developed an adaptive local fitting strategy to estimate scatter for new radiographs.
• Compared local and global fitting of various scatter models, including deep learning approaches.

Main Results:

• Local fitting of simple scatter models achieved state-of-the-art descattering performance.
• The proposed local approach reduced density reconstruction errors caused by scatter by over 50%.
• Local fitting demonstrated superior performance compared to global fitting methods.

Conclusions:

• Adaptive local modeling offers a powerful and effective strategy for scatter correction in X-ray CT.
• This method significantly enhances the quantitative accuracy of CT imaging.
• The approach provides a robust solution for descattering challenges in polyenergetic X-ray CT.