Quantitative comparison of direct phase retrieval algorithms in in-line phase tomography

Area of Science:

• Physics
• Materials Science
• Medical Imaging

Background:

• X-ray microcomputed tomography (micro-CT) faces limitations in sensitivity, particularly for soft tissues and subtle density variations.
• Phase contrast imaging, especially propagation-based in-line methods using synchrotron sources, offers a substantial sensitivity increase (up to 10^3).
• Phase retrieval is crucial for reconstructing the object's refractive index from phase shifts, enabling detailed imaging.

Purpose of the Study:

• To compare the performance of four distinct phase retrieval algorithms for propagation-based phase contrast imaging.
• To evaluate algorithms based on the transport of intensity equation (TIE), contrast transfer function (CTF), and a hybrid approach.
• To assess algorithm efficacy using both simulated and experimental data from a third-generation synchrotron source.

Main Methods:

• Implementation and comparison of four phase retrieval algorithms: TIE, TIE for weak absorption, CTF, and a mixed CTF-TIE approach.
• Linearization of the phase shift-intensity relationship for fast algorithm development.
• Acquisition of experimental data at the European Synchrotron Radiation Facility (ESRF).
• Evaluation using two quantitative reconstruction error metrics.

Main Results:

• All tested algorithms rely on linearization for computational efficiency.
• The mixed CTF-TIE approach demonstrated the best performance based on the selected reconstruction error metrics.
• While slightly less computationally efficient, the mixed approach provided superior accuracy in phase retrieval.

Conclusions:

• Propagation-based phase contrast imaging is a powerful technique for enhancing micro-CT sensitivity.
• The developed mixed phase retrieval algorithm offers improved accuracy for reconstructing object properties.
• This advancement holds promise for more detailed imaging of soft tissues and density variations in various scientific fields.