Accuracy of PET partial volume correction in tracer uptake measurements with six isotopes

Soderlund et recently published a sphere-phantom study with six PET tracers, concluding that image quality was good but tracer concentration measurements based on mean uptake (MEAN) were strongly affected by partial volume errors (PVE). In this work we ask if a previously published PVE correction method (CORR, described in 2002 by Hickeson and colleagues) is more accurate. The CORR method may be appropriate in applications like lung-nodule PET/CT when isolated hot structures, perhaps lung nodules, sit in a region of approximately uniform background, and the object's volume (V) is known from CT. A volume of interest in the PET image (VOI) is defined around the object. Background is estimated from voxels adjacent to the VOI. Tracer concentration in Bq/ml is set to the background-corrected VOI total, divided by V. Hot spheres had diameters 10, 13, 17 and 22 mm. Cold spheres had diameters 28 and 37 mm. The phantom was filled with positron-emitting
¹⁸
F,
¹¹
C,
⁸⁹
Zr,
¹²⁴
I,
⁶⁸
Ga and
⁹⁰
Y activity with a sphere to background ratio of approximately 4:1. TOF PET data from a Siemens mCT PET/CT scanner were reconstructed with the TOF-PSF method. Using the MEAN and CORR methods, we calculated contrast recovery (CR), defined by NEMA NU 2-2007, for hot and cold spheres. In hot spheres 13 mm or more in diameter, CR based on CORR was between 90% and 105% for all but one measurement with
¹⁸
F,
¹¹
C,
⁸⁹
Zr,
¹²⁴
I,
⁶⁸
Ga. In the 10 mm sphere, CR was 82% on average. In these spheres and these isotopes, CR based on MEAN varied between 0.3 and 0.8. In the cold spheres, recovery based on CORR was between 0.7 and 1.0 for these isotopes. Reconstructions of
⁹⁰
Y were less accurate. In conclusion, CORR improved contrast recovery measurements, compared to MEAN.