X-ray CT geometrical calibration via locally linear embedding

Area of Science:

• Medical Imaging
• Image Reconstruction
• Computational Imaging

Background:

• Geometric calibration and patient motion are critical for high-quality X-ray computed tomography (CT) imaging.
• Inaccurate system geometry and patient movement during scans introduce artifacts like streaks and blurring.
• Existing methods may not fully address the inter-related nature of calibration and motion compensation.

Purpose of the Study:

• To develop a novel, generalized calibration approach for X-ray CT systems.
• To address the challenge of image artifacts arising from uncalibrated geometry and rigid patient motion.
• To improve the optimization of image reconstruction quality in CT.

Main Methods:

• A locally linear embedding (LLE) based calibration method is proposed.
• Projections are represented by up-sampled neighbors using LLE.
• CT system parameters are iteratively estimated directly from projection data under a rigid 2D object assumption.

Main Results:

• The proposed method effectively calibrates CT system parameters.
• Images reconstructed using the calibrated parameters show excellent agreement with those reconstructed using true parameters.
• The approach demonstrates robustness in numerical and experimental studies.

Conclusions:

• The LLE-based calibration method offers a more general and effective solution for CT system calibration and motion compensation.
• This technique significantly reduces artifacts, leading to improved CT image quality.
• The findings have implications for enhancing diagnostic accuracy in medical imaging.