A comparative study of information retrieval in grating-based x-ray phase-contrast imaging

Area of Science:

• Medical Imaging
• X-ray Physics
• Computational Imaging

Background:

• Cosine-model analysis (CMA) and small angle x-ray scattering (SAXS) are key algorithms in grating interferometer-based X-ray phase-contrast imaging.
• Existing literature has not fully elucidated the intrinsic relationship between CMA and SAXS.

Purpose of the Study:

• To theoretically derive and experimentally verify the intrinsic connections between CMA and SAXS.
• To introduce principal component analysis (PCA) for simplified and efficient image retrieval in X-ray imaging.

Main Methods:

• Theoretical derivation and experimental verification of CMA-SAXS relationships.
• Application of deconvolution and Fourier transform within CMA.
• Development of two PCA-based algorithms for SAXS deconvolution and CMA signal extraction.
• Quantitative structural similarity (SSIM) index and profile analysis for image quality assessment.

Main Results:

• SAXS can be effectively interpreted by CMA's cosine-model assumption.
• CMA with an added deconvolution process demonstrated valid scattering distribution analysis.
• PCA-based methods showed comparable performance to Fourier transform and improved SAXS efficiency and stability.
• Retrieved images from CMA and SAXS were highly consistent, with SSIM values indicating excellent agreement.

Conclusions:

• The intrinsic link between CMA and SAXS is established, validating the cosine-model assumption for scattering distribution.
• PCA-based algorithms provide efficient, stable, and high-quality image retrieval in X-ray phase-contrast imaging.
• The simplified SAXS-analogous method offers computational advantages while maintaining image quality.