Beam-hardening correction in clinical x-ray dark-field chest radiography using deep-learning-based bone segmentation

Area of Science:

• Medical Imaging
• Radiography
• Pulmonary Diagnostics

Background:

• Dark-field radiography offers advanced lung microstructural imaging.
• It uses a Talbot-Lau interferometer for simultaneous attenuation and dark-field images.
• Polychromatic X-rays cause beam hardening, creating artifacts from bones like ribs and clavicles.

Purpose of the Study:

• To reduce bone-induced artifacts in dark-field chest radiography.
• To improve the reliability of clinical dark-field imaging.
• To suppress artificial dark-field signals from beam hardening.

Main Methods:

• Developed a segmentation-based beam-hardening correction (BHC) using deep learning.
• Utilized dual-layer detector CT data for attenuation-contribution masks.
• Trained models on chest radiographs and applied them to dark-field images and spectral CT scans.

Main Results:

• Significantly reduced bone-induced artifacts in dark-field images.
• Enhanced the homogeneity of the lung dark-field signal.
• Diminished cross-talk between attenuation and dark-field channels, improving interpretation.

Conclusions:

• The BHC method effectively suppresses artificial dark-field signals from polychromatic X-rays.
• Combines deep learning segmentation with material-specific weighting.
• Enables more reliable assessment of pulmonary microstructure in clinical settings.