Importance of algorithms in SUV calculation. Quantitative comparison of differences in SUV obtained from OSEM and Q.Clear® algorithms

M. AGOLTI; B. MOGLIA; J. A. BIURRUN MANRESA; J. BUSTOS

Vienna

Congreso; Annual Congress of the European Association of Nuclear Medicine; 2017

European Association of Nuclear Medicine

Aim: the standardized uptake value (SUV) is an objective measurement of FDG uptake, which is important in the evaluation of response to treatment. Using a new reconstruction algorithm (Q.Clear®), we noticed large differences in the resulting SUV values compared to the conventional OSEM algorithm. In this study we quantified these differences in order to assess their potential impact on treatment evaluation. Materials and methods: 72 FDG PET-CT scans performed with a GE Discovery PET/CT 710 were included in the study, acquired in our center between December 2016 and April 2017. Scans were assessed simultaneously using the same ROI size at the same location. Maximum SUV from both algorithms were obtained from the liver as a control (n=72), from lesions of 1 cm (n=28) and from lesions of 2 cm (n=28). Data was analyzed using Bland-Altman analysis, from which the mean difference and the 95% limits of agreement (LoA) were derived. The mean difference is a measure of systematic error, and the LoA represent the maximum differences due to random error between methods, within which 95% of the differences between measurements are expected to lie. Differences were calculated as (SUVQ.Clear - SUVOSEM).Results: In average, SUVQ.Clear was larger than SUVOSEM, but not in all individual cases. Mean differences in normal liver SUV were 0.1%, and the maximum differences ranged between -5.2% and 5.4%. Mean differences in SUV from 1 cm lesions were 7.5%, and the maximum differences ranged between -19.9% and 35.0%. Finally, mean differences in SUV from 2 cm lesions were 3.5%, and the maximum differences ranged between -6.0% and 13.0%.Conclusion: the differences between methods were acceptable for liver SUVs, but not for SUVs quantified from lesions. Differences were larger for smaller lesions, and the large systematic error (up to 7.5%) and random variation found (up to 35%) are clinically significant, since they could erroneously lead to a change in treatment for some patients. Therefore, SUVs obtained from different algorithms cannot be used interchangeably. These results stress the importance of visual assessment to evaluate response to treatment, since SUV is neither silly nor smart, but is rather a Simple Useful Value that has to be considered within the proper clinical context.