Tomographic image reconstruction by eigenvector decomposition: its limitations and areas of applicability

Area of Science:

• Medical Imaging
• Computational Physics

Background:

• Tomographic image reconstruction in Positron Emission Tomography (PET) systems often involves solving ill-posed inverse problems.
• The blurring matrix in PET imaging can lead to ill-conditioned systems, making accurate reconstruction difficult.

Purpose of the Study:

• To analyze the pseudo-inversion method for tomographic image reconstruction in PET.
• To investigate the physical reasons for the ill-conditioned nature of PET reconstruction problems.
• To evaluate the utility of pseudo-inversion for specific PET configurations and the inclusion of time-of-flight (TOF) data.

Main Methods:

• Utilizing eigenvector and eigenvalue decomposition to assess matrix conditioning.
• Applying pseudo-inversion techniques to the blurring matrix of a PET system.
• Analyzing planar camera configurations and incorporating time-of-flight (TOF) information.

Main Results:

• Finding an accurate pseudo-inverse is computationally challenging for standard PET arrays due to ill-conditioning.
• Pseudo-inversion proves useful for incomplete detector rings and non-invariant point response functions.
• Planar camera configurations result in well-conditioned, separable matrices, allowing for smaller pixel sizes with controlled noise magnification.

Conclusions:

• Despite computational challenges, pseudo-inversion is a valuable reconstruction method for specific PET scenarios, particularly planar imaging.
• Incorporating time-of-flight (TOF) information can further enhance the reconstruction process for PET imaging.