Calibration of PET/CT scanners for multicenter studies on differentiated thyroid cancer with 124I

Jakob W. Kist, Manfred van der Vlies, Otto S. Hoekstra, Henri N. J. M. Greuter, Bart de Keizer, Marcel P. M. Stokkel, Wouter V. Vogel, Marc C. Huisman, Arthur van Lingen

Studies on imaging of differentiated thyroid cancer (DTC) using (124)I often require a multicenter approach, as the prevalence of DTC is low. Calibration of participating scanners is required to obtain comparable quantification. As determination of a well-defined range of recovery coefficients is complicated for various reasons, a simpler approach based on the assumption that the iodine uptake is highly focal with a background that significantly lacks radioactivity might be more efficient. For each scanner, a linear conversion between known and observed activity can be derived, allowing quantification that can be traced to a common source for all scanners within one study-protocol. The aim of this paper is to outline a procedure using this approach in order to set up a multicenter calibration of PET/CT scanners for (124)I. A cylindrical polyethylene phantom contained six 2-ml vials with reference activities of ~2, 10, 20, 100, 400, and 2000 kBq, produced by dilution from a known activity. The phantom was scanned twice on PET/CT scanners of participating centers within 1 week. For each scanner, the best proportional and linear fit between measured and known activities were derived and based on statistical analyses of the results of all scanners; it was determined which fit should be applied. In addition, a Bland-Altman analysis was done on calibrated activities with respect to reference activities to asses the relative precision of the scanners. Nine Philips (vendor A) and nine Siemens (vendor B) PET/CT scanners were calibrated in a time period of 3 days before and after the reference time. No significant differences were detected between the two subsequent scans on any scanner. Six fitted intercepts of vendor A were significantly different from zero, so the linear model was used. Intercepts ranged from -8 to 26 kBq and slopes ranged from 0.80 to 0.98. Bland-Altman analysis of calibrated and reference activities showed that the relative error of calibrated activities was smaller than that of uncalibrated activities. A simplified multicenter calibration procedure for PET/CT scans that show highly focal uptake and negligible background is feasible and results in more precise quantification. Our procedure can be used in multicenter (124)I PET scans focusing on (recurrent) DTC.