Improved kinetic analysis of dynamic PET data with optimized HYPR-LR

Area of Science:

• Medical Imaging
• Nuclear Medicine
• Image Reconstruction

Background:

• Highly constrained backprojection-local reconstruction (HYPR-LR) enhances signal-to-noise ratio (SNR) in dynamic imaging, showing promise for positron emission tomography (PET).
• HYPR-LR can improve kinetic analysis methods in PET that are sensitive to noise, potentially leading to more accurate parameter estimates.
• Existing HYPR-LR methods may introduce bias into kinetic parameter estimates in non-sparse PET studies.

Purpose of the Study:

• To examine the performance of HYPR-LR in PET kinetic analysis.
• To develop a tailored HYPR-LR implementation (HYPR-LR-MC) to minimize bias and maximize variance reduction.
• To provide a framework for validating HYPR-LR processing for specific PET imaging tasks.

Main Methods:

• Developed HYPR-LR-MC using multiple temporally summed composite images based on tracer kinetics.
• Compared HYPR-LR-MC and HYPR-LR with a full composite (HYPR-LR-FC) using [11C]-PIB PET data.
• Evaluated performance using simulated and human studies, generating nondisplaceable binding potential (BP(ND)) parametric images with Logan analysis and receptor parametric mapping (RPM2).

Main Results:

• HYPR-LR-FC overestimated BP(ND) in high uptake regions; HYPR-LR-MC virtually eliminated this bias.
• Both HYPR-LR methods reduced variance in parametric images, with HYPR-LR-FC offering greater reduction.
• HYPR-LR processing favorably compared to spatial smoothing, reducing variance without loss of spatial resolution.

Conclusions:

• HYPR-LR processing significantly reduces variance and eliminates noise-dependent bias in PET parametric images.
• HYPR-LR-MC effectively removes bias associated with high-uptake regions, albeit with slightly less variance reduction than HYPR-LR-FC.
• The proposed HYPR-LR-MC method offers a validated approach for improving PET kinetic analysis.