Accuracy of propagation-based phase-contrast CT under the projection approximation

Abstract

X-ray phase-contrast imaging has the potential to improve image contrast with lower dose by probing an object's refractive properties as well as its absorptive properties. To reconstruct a phase-contrast image from a raw dataset, a phase retrieval algorithm must be applied to invert the forward model of the image acquisition scheme. Discrete forward modeling presents unique computational challenges due to high x-ray wave field sampling requirements. At the cost of accuracy, approximations are often applied for the sake of simplicity and experimental convenience. One of the most ubiquitous simplifications is the projection approximation, which neglects refractive effects within an object. The approximation's validity decreases when imaging thicker objects or using a detector with higher spatial resolution. For greater accuracy, one might use the multislice model instead. In this work, we explore the accuracy of the projection approximation in simulated synchrotron micro-CT images. We simulated images using two detector resolutions (0.5 and 2 micron) with either the projection approximation or multislice forward model. Paganin phase retrieval was applied to the resulting datasets, and the final phase-contrast images were reconstructed using filtered back-projection. The 2-micron resolution detector images showed little difference between forward models, indicating the projection approximation is satisfied. With sub-micron resolution, there is a noticeable difference in the phase-contrast images, especially around fine details. An accurate phase retrieval algorithm for such high-resolution data likely requires more detailed forward modeling that avoids making the projection approximation.