Physical phantom validation of clustering-initiated factorization in dynamic PET

Area of Science:

• Neuroimaging
• Radiochemistry
• Biophysics

Background:

• Dynamic positron emission tomography (PET) quantifies physiological parameters for neuropsychiatric disorder research.
• Cluster-Initialized Factor Analysis (CIFA) is a factor analysis algorithm that overcomes reference region limitations.
• CIFA automatically extracts radiotracer binding distributions by analyzing tracer dynamics, distinguishing specific and non-specific binding without prior segmentation.

Purpose of the Study:

• To quantitatively validate CIFA's ability to resolve dynamic biological processes using an independent benchmark.
• To develop a physical phantom capable of modeling unique aspects of dynamic PET imaging for benchmark evaluation.

Main Methods:

• CIFA was applied to reconstruct 18F-flortaucipir dynamic brain PET datasets.
• A custom physical brain phantom with hydraulic elements simulated overlapping tissues and free radiotracer.
• The phantom generated dynamic scans with realistic partial volume effects and varying time-activity curve (TAC) similarities for 10 distinct simulations.

Main Results:

• CIFA performance was evaluated by correlating estimated outputs with ground truth tissue TACs and distributions.
• For seven out of 10 modeled dynamics, the curve correlation for specific binding tissue exceeded 95%.
• These results represent a full spectrum of realistically expected tissue TAC shapes.

Conclusions:

• An innovative process combining a physical phantom and PET images was developed to evaluate CIFA.
• CIFA accurately reproduced the dynamics of simulated data from the physical phantom in most cases.
• This validates CIFA's application in extracting dynamic TACs from dynamic PET imaging data.