Development of an adjustable dynamic phantom for testing deviceless motion correction algorithm in PET

Respiratory motion during PET Imaging leads to blurring, volumetric overestimation, and reduced quantitative accuracy. Deviceless motion correction algorithms (MCA) such as OncoFreeze (OF) by Siemens Healthineers (Forchheim, Germany) extract respiratory waveforms from list-mode data but require validation that cannot be obtained in vivo.

To develop and validate an adjustable dynamic phantom with fillable spheres for quantitative assessment of deviceless PET motion correction.

A microprocessor-controlled two-axis phantom was constructed to generate linear craniocaudal (cc), ventrodorsal (vd), and diagonal (diag) motion with amplitudes of 0-40 mm and frequencies of 10-40 min^{- 1}. Polymethylmethacrylate (PMMA) spheres (4, 1, and 0.25 mL) filled with ^{1 8}F-fluorodeoxyglucose (FDG) were placed in a water-filled PMMA tank and imaged on a Biograph Vision 600 PET/CT (Siemens Healthineers AG, Forchheim, Germany). For each sphere and direction, 2-min PET scans were acquired with and without MCA, yielding 288 datasets. Volumes, activities, and recovery coefficients (RCs) were extracted using two different isocontours.

Without correction, motion produced substantial overestimation of apparent volume: +15%-55% (4 mL), +81%-166% (1 mL) and +276%-509% (0.25 mL), depending on trajectory and IC₃₇. MCA reduced these distortions: For the 1 mL sphere from +87% to -8% (vd) and +166% to +82% (diag). IC_0.1-based activity remained within ≈1%-2% of the nominal value and was minimally affected by MCA, while IC₃₇ underestimated activity (∼45%) and was direction-dependently altered by correction. RCs increased with sphere size and were partially improved by MCA but not restored to static levels.

The phantom provides a reproducible ground-truth framework for evaluating deviceless PET motion correction. MCA reduces motion-induced blurring and volumetric inflation, though correction remains size- and direction-dependent. The system enables systematic benchmarking and further optimization of motion-correction algorithms.