Comparison of methods for sensitivity correction in Talbot-Lau computed tomography

Area of Science:

• Medical Imaging
• Computational Imaging
• X-ray Optics

Background:

• Talbot-Lau X-ray phase contrast imaging is sensitive to object positioning.
• Large objects can cause inhomogeneous phase contributions, leading to artifacts in tomographic reconstructions.

Purpose of the Study:

• To compare and evaluate recently proposed methods for correcting reconstruction artifacts in Talbot-Lau X-ray phase contrast imaging.
• To assess the performance of iterative reconstruction, known operator networks, and U-nets for artifact reduction.

Main Methods:

• Qualitative and quantitative comparison of three artifact correction methods: iterative reconstruction, known operator network, and U-net.
• Evaluation using the Shepp-Logan phantom and human abdomen anatomy data.
• Dedicated experiments to analyze the noise behavior of each method.

Main Results:

• All evaluated methods successfully reduced artifacts in simulated and real anatomy data.
• Method-specific residual errors were observed, reflecting distinct correction strategies.
• Differential noise behavior was noted across the compared artifact correction techniques.

Conclusions:

• Iterative reconstruction offers high performance but requires significant runtime.
• Known operator networks demonstrate consistently competitive performance.
• U-nets provide a general-purpose solution with slightly lower performance but broader applicability.