AB-OSEM reconstruction for improved Patlak kinetic parameter estimation: a simulation study.

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Image Reconstruction

Background:

• OSEM reconstruction introduces positive bias in cold regions, particularly with low-count data common in dynamic PET imaging.
• The non-negativity constraint, while reducing standard deviation, can be detrimental to quantitative accuracy.

Purpose of the Study:

• To investigate algorithms that relax the non-negativity constraint in image reconstruction.
• To compare the performance of AB-OSEM and NEG-ML algorithms against standard OSEM and FBP.
• To evaluate reconstruction methods for static and dynamic PET imaging, including kinetic parameter estimation.

Main Methods:

• Investigated two non-negativity-constrained algorithms: NEG-ML and AB-OSEM.
• Developed AB-OSEM to incorporate randoms and scatter background terms.
• Conducted simulation studies comparing OSEM, FBP, and AB-OSEM for static and dynamic PET data.
• Assessed reconstruction performance based on bias, standard deviation, and root mean squared error (RMSE).

Main Results:

• AB-OSEM outperformed NEG-ML reconstruction.
• AB-OSEM, with a negative lower bound, successfully avoided the positive bias seen in OSEM.
• AB-OSEM demonstrated lower overall RMSE for activity concentration and kinetic parameters compared to OSEM and FBP.
• Parametric images from high-resolution research tomograph were generated using different reconstruction methods.

Conclusions:

• Lifting the non-negativity constraint using AB-OSEM improves quantitative accuracy in PET imaging.
• AB-OSEM is a preferred reconstruction method over standard OSEM and FBP for kinetic parameter estimation from dynamic PET data.
• The AB-OSEM algorithm offers a robust approach for reducing errors in both static and dynamic PET quantification.