Program Information
Impact of PET Scanner Harmonization On Quantitative Response Assessment
A Weisman1*, T Bradshaw1 , C Lin1 , R Jeraj1,2 , (1) University of Wisconsin, Madison, WI, (2) University of Ljubljana, Ljubljana, Slovenia
Presentations
SU-F-605-6 (Sunday, July 30, 2017) 2:05 PM - 3:00 PM Room: 605
Purpose: Patients with solid tumors often receive [F-18]-FDG PET/CT scans on different scanners throughout monitoring period. In this study, we quantified the impact of scanner harmonization on longitudinal PET changes and determined patient and lesion-specific factors that most influenced harmonization.
Methods: Twenty metastatic melanoma patients received at least two FDG PET/CT scans, each on two different scanners with an average of 4 months (range: 2-8) between. Scans from a GE Discovery 710 PET/CT were harmonized to the GE Discovery VCT using image reconstruction settings matching recovery coefficients and noise index in phantoms. SUVmax and SUVpeak were measured for each lesion and time-point. H, defined as H=abs((∆SUVorig–∆SUVharm)/SUVbaseline), captured the impact of harmonization on longitudinal relative changes. Multivariate linear mixed-effects regression was performed between H and patient/lesion factors (e.g. patient weight, absolute lesion SUV). Significant longitudinal changes in SUV were determined using limits of agreement from literature.
Results: Ninety observations were available for harmonization assessment. Median H was 31% for SUVmax (lower-upper quartiles: 18-49%) and 20% for SUVpeak (12-27%). The impact of harmonization was greater for lesions with higher SUVmax (p<0.001, 3% increase in H per unit SUVmax). Lesions located further from machine isocenter were more sensitive to harmonization, likely due to time-of-flight and point spread function modeling (p=0.046, 5% increase in H per cm distance increase). In 33 observations (37%), significant changes in SUVmax on original reconstructions were no longer significant after harmonization or vice versa. This occurred in 25 observations (28%) for SUVpeak.
Conclusion: Longitudinal change in SUVmax was shown to be highly sensitive to scanner harmonization, especially for higher intensity lesions, while SUVpeak was more robust. Harmonization impacted lesion response interpretation in about 1/3 of observations, indicating caution must be taken when assessing quantitative PET changes on different scanners without harmonization.
Contact Email: