Sinogram Blurring Matrix Estimation From Point Sources Measurements With Rank-One Approximation for Fully 3-D PET

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Image Reconstruction

Background:

• Accurate system matrices are crucial for high-quality positron emission tomography (PET) image reconstruction.
• Existing methods using 2-D blurring matrices do not fully compensate for inter-sinogram blurring, impacting image quality in long axial field-of-view scanners.
• 4-D blurring matrix estimation is desirable but faces challenges due to the ill-conditioned nature of the problem.

Purpose of the Study:

• To develop a stable and efficient method for estimating a 4-D blurring matrix for PET scanners.
• To improve image quality by compensating for both transaxial and inter-sinogram blurring effects.
• To reduce the number of point source scans required for accurate system matrix estimation.

Main Methods:

• Factoring the system matrix into geometry projection and sinogram blurring matrices.
• Computing the geometry projection matrix analytically.
• Estimating a 4-D sinogram blurring matrix using a novel rank-one approximation for improved stability.
• Applying the method to simulated and real data from an Inveon microPET scanner.

Main Results:

• The proposed rank-one approximation enhances the stability of 4-D blurring matrix estimation.
• The 4-D blurring matrix significantly improves image quality compared to a 2-D blurring matrix.
• The method achieves comparable image quality with fewer point source scans than unconstrained 4-D estimation.

Conclusions:

• The rank-one approximation method provides a stable approach for 4-D blurring matrix estimation in PET.
• This technique effectively addresses inter-sinogram blurring, leading to superior image quality.
• The optimized estimation reduces data acquisition requirements, making PET imaging more efficient.