Robustness of phase retrieval methods in X-ray phase contrast imaging: a comparison

Area of Science:

• Medical Imaging
• Computational Imaging
• Phase Retrieval Algorithms

Background:

• Robust phase retrieval is essential for reducing radiation dose in X-ray phase contrast imaging.
• Current methods face challenges with noise and singularities, impacting image quality and accuracy.

Purpose of the Study:

• To compare the robustness of two phase retrieval methods: Transport of Intensity Equation (TIE) and Attenuation-Partition based (AP-based).
• To evaluate their performance in reconstructing phase maps from experimental X-ray projection images of a phantom.

Main Methods:

• Applied TIE-based and AP-based phase retrieval methods to experimental projection images of an air-bubble wrap phantom.
• Analyzed and compared the resulting phase maps for clarity and accuracy of recovered features.

Main Results:

• The TIE-based method failed to recover bubble structures, producing erroneous phase values even with Tikhonov regularization.
• The AP-based method successfully recovered bubble details with phase values close to thickness-based estimates.
• Performance differences were attributed to distinct handling of singularity problems in phase retrieval.

Conclusions:

• The conventional TIE-based method is unstable with noisy projection images.
• The AP-based phase retrieval method demonstrates superior robustness and accuracy compared to the TIE-based method for X-ray phase contrast imaging.