Phantom-based evaluation of quantification accuracy of combined PET/MRI

Ziel/Aim:

A widely discussed question in PET is the quantitative accuracy (QA) of combined PET/MRI systems. To evaluate QA of the Philips Ingenuity TF PET/MR system we have performed phantom-based evaluations and compared them to measurements on traditional PET systems.

Methodik/Methods:

With the Philips PET/MR a SUV calibration is the standard procedure to calibrate the PET against a dose calibrator (DC). A dynamic scan is started at high activity (≈500 MBq) and run for 7 half-life times. Than a SUV validation with lower activity (≈70 MBq) is run as a static scan for 3 min. We augmented the validation protocol to start at higher activity and to run as a dynamic protocol. To evaluate the clinical relevant QA we have determined typical singles rates (SR) for whole-body (WB) scans. To analyse the influence of the MR-based attenuation (MRAC) we have created an attenuation template (AT) of the phantom and applied it.

Ergebnisse/Results:

The PET-derived activity concentrations and SUVs are significantly lower than calculated from the DC. Using the standard procedure the PET vs. DC ratio is 0.895=10.5% deviation at a level of 73.7 MBq (SR: 14.7 Mcps). For activity increasing from 8.6 to 89.7 MBq the PET/DC ratio decreases from 0.943 to 0.877, corresponding to deviations of 6–12% (SR: 3.8–17.0 Mcps). The use of an AT changes the deviation < 1%. For typical WB scans the SR were found to range from 8.0–16.6 Mcps, corresponding to a underestimates of true activity by 6.8–9.9%.

Schlussfolgerungen/Conclusions:

Our preliminary results show, that QA of the Philips PET/MR in phantoms is off by 6–12%. The effect shows a notable dead-time dependency and cannot be explained by inaccurate MRAC. In addition to the dead-time related variance of ≈3% there remains a systematic error in the calibration, which biases QA of clinical scans by ≈6%. Philips has been informed about the inconsistencies and has acknowledged to provide a fix.