Comparison of segmentation methods for evaluating liver uptake of 18F-FDG in PET/CT images

Abstract

Positron emission tomography combined with computed tomography (PET/CT) using the radiopharmaceutical 18F-FDG is widely employed in oncology, with the liver often used as an internal reference for quantifying metabolic uptake, including the Metabolic Tumor Volume (MTV). Radiotracer uptake is expressed by the Standardized Uptake Value (SUV), whose determination is influenced by the segmentation method adopted. This study aimed to compare three segmentation methods for assessing hepatic SUV in PET/CT scans of 74 patients with multiple myeloma: (i) a fixed spherical volume of interest (VOI), 2 cm in radius, positioned in the right liver lobe, (ii) auto-segmentation by CT image applying a global threshold with manual adjustments (AS), and (iii) segmentation by artificial intelligence (AI) using the TotalSegmentator software. Images of patients seen at the Nuclear Medicine Service of Unicamp between 2013 and 2018 were retrospectively analyzed. Statistical comparison among methods was performed using the Kruskal–Wallis test, followed by Dunn’s post hoc test. The results showed that mean SUV did not present significant differences among the three methods. However, for maximum SUV and peak SUV, statistically significant differences were observed in relation to the spherical VOI, while AS and AI did not differ from each other. Furthermore, a discrepancy of approximately 29.3% was observed between the volumes segmented by AS and AI. These findings suggest that, although more complex methods allow segmentation of the entire liver, the spherical VOI represents a simple and reliable alternative for obtaining the mean SUV, a parameter widely used in clinical analyses and research.