Optimization of Dark-Field CT for Lung Imaging

TL;DR

This paper presents an optimization method for dark-field CT lung imaging, focusing on system parameters to enhance contrast-to-noise ratio, and provides design principles for improved system performance.

Contribution

It introduces a CNR-based metric for optimizing dark-field CT system parameters and analyzes the relationship between system geometry, material properties, and optimal auto-correlation length.

Findings

Optimal auto-correlation length (ACL) is about 0.21 μm for the phantom.

Larger source-detector distance increases maximum ACL, aiding system optimization.

Optimal ACL depends mainly on phantom scattering material's size and absorption.

Abstract

Background: X-ray grating-based dark-field imaging can sense the small angle scattering caused by an object's micro-structure. This technique is sensitive to lung's porous alveoli and is able to detect lung disease at an early stage. Up to now, a human-scale dark-field CT has been built for lung imaging. Purpose: This study aimed to develop a more thorough optimization method for dark-field lung CT and summarize principles for system design. Methods: We proposed a metric in the form of contrast-to-noise ratio (CNR) for system parameter optimization, and designed a phantom with concentric circle shape to fit the task of lung disease detection. Finally, we developed the calculation method of the CNR metric, and analyzed the relation between CNR and system parameters. Results: We showed that with other parameters held constant, the CNR first increases and then decreases with the system auto-correlation length (ACL). The optimal ACL is nearly not influenced by system's visibility, and is only related to phantom's property, i.e., scattering material's size and phantom's absorption. For our phantom, the optimal ACL is about 0.21 {\mu}m. As for system geometry, larger source-detector and isocenter-detector distance can increase the system's maximal ACL, helping the system meet the optimal ACL more easily. Conclusions: This study proposed a more reasonable metric and a task-based process for optimization, and demonstrated that the system optimal ACL is only related to the phantom's property.