4D-PET reconstruction using a spline-residue model with spatial and temporal roughness penalties

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Image Reconstruction

Background:

• Dynamic positron emission tomography (dPET) 4D reconstruction enhances image quality by fitting temporal functions to voxel time-activity curves (TACs).
• The optimal temporal function for dPET reconstruction remains an open research question, impacting noise suppression and bias.
• Accurate TAC modeling is crucial for quantitative analysis in dPET, particularly for oncology applications.

Purpose of the Study:

• To introduce and evaluate a novel spline-residue model for 4D dPET reconstruction.
• To compare the performance of the spline-residue model against conventional and other advanced 4D reconstruction methods.
• To assess the impact of temporal regularization on dPET reconstruction accuracy and noise.

Main Methods:

• Developed a spline-residue model describing TACs via convolutions of the arterial input function with cubic B-spline basis functions.
• Implemented 4D reconstructions using a nested-MAP algorithm with spatial and temporal roughness penalties.
• Tested algorithms with Monte Carlo simulated dPET data for a thoracic phantom, mimicking [18F]-Fluromisonidazole kinetics in lung cancer.

Main Results:

• Spline-residue 4D reconstruction demonstrated >50% improvement in bias and noise for 5/8 kinetic parameters compared to conventional MAP reconstruction.
• The spline-residue model outperformed other 4D methods in 5/8 combinations, showing superior noise suppression and bias reduction.
• Incorporating temporal roughness penalties enhanced the performance of spline-residue, spectral, and B-spline models.

Conclusions:

• The spline-residue model offers a robust approach for 4D dPET reconstruction, improving quantitative accuracy and image quality.
• This method provides a significant advancement over traditional reconstruction techniques for dynamic PET imaging.
• Further development and application of the spline-residue model hold promise for enhanced diagnostic capabilities in [18F]-Fluromisonidazole PET imaging and other dPET applications.