Multi-centre calibration of an adaptive thresholding method for PET-based delineation of tumour volumes in radiotherapy planning of lung cancer

Purpose: To evaluate the calibration of an adaptive thresholding algorithm (contrast-oriented algorithm) for FOG PET-based delineation of tumour volumes in eleven centres with respect to scanner types and image data processing by phantom measurements. Methods:A cylindrical phantom with spheres of different diameters was filled with FDG realizing different signal-to-background ratios and scanned using 5 Siemens Biograph PET/CT scanners, 5 Philips Gemini PET/CT scanners, and one Siemens ECAT-ART PET scanner. All scans were analysed by the contrast-oriented algorithm implemented in two different software packages. For each site, the threshold SUVs of all spheres best matching the known sphere volumes were determined. Calibration parameters a and b were calculated for each combination of scanner and image-analysis software package. In addition, scanner-type-specific calibration curves were determined from all values obtained for each combination of scanner type and software package. Both kinds of calibration curves were used for volume delineation of the spheres. Results: Only minor differences in calibration parameters were observed for scanners of the same type (Delta a <= 4%, Delta b <= 14%) provided that identical imaging protocols were used whereas significant differences were found comparing calibration parameters of the ART scanner with those of scanners of different type (Delta a <= 60%, Delta b <= 54%). After calibration, for all scanners investigated the calculated SUV thresholds for auto-contouring did not differ significantly (all p > 0.58). The resulting sphere volumes deviated by less than -7% to +8% from the true values. Conclusion: After multi-centre calibration the use of the contrast-oriented algorithm for FDG PET-based delineation of tumour volumes in the different centres using different scanner types and specific imaging protocols is feasible.