Evaluation of a direct 4D reconstruction method using generalised linear least squares for estimating nonlinear...

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Biophysics

Background:

• Estimating nonlinear micro-parameters in medical imaging is computationally intensive and prone to noise.
• Conventional methods often yield noisy parametric maps and are time-consuming.

Purpose of the Study:

• To apply a direct parametric image reconstruction algorithm for estimating micro-parameters of a two-tissue compartment model for [18F]FDG kinetics.
• To improve the accuracy and efficiency of parametric map generation in dynamic PET imaging.

Main Methods:

• Employed a direct parametric reconstruction algorithm that separates tomographic and kinetic modeling.
• Utilized the generalized linear least squares (GLLS) algorithm for post-reconstruction analysis.
• Validated the method on both clinical and simulated dynamic PET data.

Main Results:

• The direct reconstruction method demonstrated significant quantitative and qualitative improvements in micro-parameter estimation compared to conventional methods.
• Parametric maps showed substantial quantitative agreement with established methods like filtered back projection and direct Patlak.
• Reduced variance in voxel-wise parametric maps was observed.

Conclusions:

• The proposed direct parametric reconstruction algorithm is a promising approach for estimating micro-parameters in compartment modeling.
• The linearised GLLS algorithm allows for efficient implementation, minimizing impact on overall reconstruction time.
• This method enhances the reliability and speed of kinetic modeling in dynamic PET imaging.