Foam-like phantoms for comparing tomography algorithms.

Area of Science:

• Computational imaging
• Image reconstruction
• Scientific computing

Background:

• Current tomographic algorithm benchmarking relies on datasets with limitations.
• Real-world datasets lack flexibility and sample size.
• Mathematical phantoms are often too simple or unrepresentative.

Purpose of the Study:

• To develop a new family of mathematical phantoms for tomographic algorithm comparison.
• To address the limitations of existing benchmark datasets.
• To enable fair and informative evaluation of tomographic reconstruction algorithms.

Main Methods:

• Generation of complex foam-like mathematical phantoms with over 100,000 features.
• Simulation of various acquisition modes and experimental conditions.
• Development of 4D extensions for dynamic tomography algorithm comparison.

Main Results:

• The proposed phantoms are challenging to reconstruct and representative of real samples.
• Computer-generated phantoms allow for simulated acquisition modes and unlimited variations.
• 4D extensions facilitate comparisons for dynamic tomography.

Conclusions:

• The novel foam-like phantoms satisfy key requirements for benchmark datasets.
• These phantoms enable fair and informative comparisons of tomographic algorithms.
• The methodology supports computationally efficient virtual tomographic experiments.