A novel method for in-vivo evaluation of quantification accuracy in PET

Ziel/Aim:

One of the strengths of PET is the ability to measure regional tracer concentrations, which is the basis of a quantitative evaluation of patient studies. Parameters such as the standardized uptake value (SUV) or transport constants derived from kinetic modeling are commonly used for this purpose. The quantitative accuracy strongly depends on accurate calibration of the scanner, which is commonly performed with suitable phantoms. While this calibration procedure ensures quantitatively correct results in geometrically comparable phantoms, it still leaves questions to be asked regarding its reliability for in-vivo measurements due to different scatter distribution as well as tracer distribution in phantoms compared to patient scans.
It was therefore the aim of our study to investigate the actual in-vivo accuracy of a phantom-based scanner cross-calibration against a well-counter by comparison of the tracer concentration in the patient’s urine as determined by the PET system to the concentration measured in a well-counter.

Methodik/Methods:

In 39 routinely scanned F18-FDG patients on two different PET scanners (EXACT HR+ PET: N=22; Biograph PET/CT: N=17) the bladder region was imaged as the last bed position and urine samples collected immediately after the end of the scan. 3D region-of-interests (ROI) were placed in the bladder by three different observers via a threshold-based delineation method using the ROVER software and image-based activity concentrations were determined. Activity concentration in the urine samples was determined in well-counters which were cross-calibrated against the respective scanner via routine phantom procedures.

Ergebnisse/Results:

Our first results show, that the measured activity concentrations are significantly lower in PET than in the well-counter for all patients and both investigated scanners; PET vs. well-counter ratio: 0.88±0.06 [range: 0.79-1.10] (EXACT HR+), 0.86±0.11 [range: 0.57-1.06] (Biograph).

Schlussfolgerungen/Conclusions:

The presented results suggest, that the commonly used phantom-based scanner calibration is not sufficient to guarantee high quantitative in-vivo accuracy. A possible reason for the observed deviations of up to 43% and the patient and scanner related variations of up to 49% might be an insufficient scatter correction in the manufacturer provided image reconstruction. To further investigate, we are currently acquiring more data sets from a third PET scanner (Gemini-TF PET/CT). We propose to augment standard calibration procedures by measurement of the in-vivo accuracy of the calibration using the described method. Furthermore, we believe that the proposed in-vivo method might also be of value for multi-center studies where monitoring the calibration accuracy for each scanner could be an important benefit and quality control method.