Impact of image acquisition and reconstruction on standardized uptake values and radiomic features of oncologic PET/CT scans

Ziel/Aim Quantitative parameters, including radiomic features, have shown promise for the analysis of oncologic PET scans. In order to use these parameters in multicenter trials and in the clinic it is necessary to understand to which extent these parameters are influenced by image reconstruction parameters, which differ between clinical sites and evolve over time at individual sites.

Methodik/Methods We evaluated the scans of the prospective multicenter clinical trial “MEMORI” which has evaluated FDG PET for monitoring tumor response to therapy in esophagogastric junction carcinomas. Tumors were delineated by 40% isocontours and standardized uptake values (SUV_mean/max), total lesion glyocolysis (TLG), skewness, kurtosis, entropy, and energy of the primary tumor were calculated using LIFEx. Images of Siemens mCT scanners, reconstructed with and without time-of-flight information and point spread function correction were analyzed. Variability was evaluated by Bland-Altman plots and coefficients of variations (COVs).

Ergebnisse/Results 62 scans from 3 centers (age 44-88 years) were analyzed. SUVs and TLG, showed a proportional bias (i.e. the differences between the two measurements increased linearly with increasing values) and were log-transformed for further analyses. Of the radiomic parameters entropy was least affected by image reconstruction (COV 4%), followed by energy (COV 17%), kurtosis (COV 20%), and skewness (COV 764%). For the log-transformed parameters the lowest variability was found for TLG (COV 2%), followed by SUV_max (COV 8%), and SUV_mean (COV 9%).

Schlussfolgerungen/Conclusions Image acquisition and reconstruction parameters had a markedly different effect on the studied radiomic features with kurtosis showing much higher variability than entropy. TLG was more stable than SUV_max and SUV_mean.

Publication History

Article published online:
14 April 2022