In-vivo evaluation of quantification accuracy of combined PET/MRI

Ziel/Aim:

Quantitative accuracy (QA) of PET parameters like standardised uptake values (SUV) and tracer kinetic rate constants strongly depends on applicability of the (phantom-based) scanner calibration in-vivo. In a previous study we introduced a method for direct in-vivo QA evaluation: for 3 PET(/CT) systems the activity concentrations (Cs) of urine samples measured in a well-counter were compared to the Cs of PET images of the bladder in 56 patients. That study showed a small (7-12%) but systematic underestimation of the true Cs by PET. To address the widely discussed question of QA of combined PET/MRI we have applied this method to our Philips PET/MR.

Methodik/Methods:

Up to now 10 clinical F18-FDG scans have been evaluated. The bladder region was imaged as the last bed position and urine samples collected afterwards. Using the ROVER software, 3D region-of-interests (ROI) of the bladder were delineated by 3 observers. To exclude partial volume effects ROIs were concentrically shrunk by 8–10 mm. Then, Cs were determined in the PET images of the bladder as well as in the urine samples using a cross-calibrated well-counter.

Ergebnisse/Results:

The measured Cs and SUVs are significantly lower in PET/MR than in the well-counter: ratio=0.773±0.035 (mean±SEM) [0.617–0.942] (p=0.00011). After correcting for known (but not yet fixed) inconsistencies in the manufacturer’s scanner calibration of the Philips PET/MR, the ratio is 0.827±0.035 [0.664–1.004] (p=0.0011).

Schlussfolgerungen/Conclusions:

Our preliminary results indicate that the Philips PET/MR underestimates true Cs by ≈17% in-vivo which is ≈7% larger than we observed for PET(/CT). The requirement to apply a correction factor due to inconsistencies in the manufacturer’s provided calibration and similar deviations reported on other PET/MR systems show that vendors of PET/MR systems need to pay special attention regarding QA. In this context, our method might serve as a convenient method to evaluate the actual QA in-vivo.