Focal spot mitigation with dynamic sampling conditions for high-resolution X-ray CT via diffusion priors

Area of Science:

• Medical Imaging
• Computational Imaging
• Image Reconstruction

Background:

• High-resolution X-ray CT imaging is crucial for clinical diagnosis.
• Focal spot dynamics significantly degrade image quality, especially in low-dose protocols.
• Current methods struggle to address these dynamic source distortions effectively.

Purpose of the Study:

• To introduce dynamic focal spot diffusion CT (DFSD-CT), a novel reconstruction framework.
• To jointly estimate the tomographic image and the time-varying focal spot profile.
• To improve image fidelity and resolution in low-dose X-ray CT.

Main Methods:

• Developed a reconstruction framework (DFSD-CT) integrating a conditional forward operator.
• Utilized a generative diffusion prior to model stochastic, time-varying focal spot behavior.
• Jointly estimated the tomographic image and the dynamic focal spot profile.

Main Results:

• DFSD-CT demonstrated improved modulation transfer function (MTF) and reconstruction fidelity by up to 3 dB.
• Achieved superior performance compared to state-of-the-art diffusion algorithms.
• Significantly narrowed the resolution gap between clinical and micro-CT imaging.

Conclusions:

• DFSD-CT effectively accounts for instantaneous and view-dependent source distortions.
• Eliminates the need for post hoc deblurring, offering a streamlined approach.
• Provides a pathway for enhanced diagnostic accuracy in low-dose X-ray CT while maintaining radiation safety.