Computer vision tools to optimize reconstruction parameters in x-ray in-line phase tomography

Area of Science:

• Medical Imaging
• Computer Vision
• Biomedical Engineering

Background:

• X-ray in-line phase tomography requires precise reconstruction parameters.
• Manual optimization relies on expert visual inspection or assumptions.
• This process can be subjective and time-consuming.

Purpose of the Study:

• To automate and improve the optimization of reconstruction parameters in x-ray in-line phase tomography.
• To introduce objective computer vision tools for parameter selection.
• To demonstrate the utility of these tools in biomedical imaging.

Main Methods:

• Utilized three computer vision tools: Scale Invariant Feature Transform (SIFT), a focus measure, and a tractography-based measure.
• Applied these tools to inject priors on object shape and scale.
• Integrated these methods with the Paganin single intensity image phase retrieval algorithm.

Main Results:

• Demonstrated the effectiveness of computer vision tools in replacing expert judgment for parameter optimization.
• Showcased the ability to incorporate object shape and scale priors.
• Successfully applied the method to heterogeneous soft tissues of biomedical interest.

Conclusions:

• Computer vision tools offer an objective and efficient alternative to manual optimization in x-ray phase tomography.
• These automated methods enhance the reliability and reproducibility of image reconstruction.
• The approach is particularly valuable for complex biomedical samples.