Efficient semi-automatic segmentation for preclinical PET/CT: faster processing without loss of accuracy
Nuklearmedizin - NuclearMedicine, vol. 65, pp. 131
Abstract
Ziel/Aim: Manual segmentation of molecular imaging scans is time-consuming and subject to operator variability, limiting throughput and reproducibility, particularly in preclinical small-animal PET/CT. This study compared a conventional manual segmentation method with a semi-automatic approach to evaluate whether the latter maintains quantitative accuracy across different voxel-inclusion thresholds. Methodik/Methods: Pre-existing PET/CT scans from several subcutaneous mouse xenograft models from our department (n=77) were included in this study. Tumor tracer uptake was quantified using both manual as well as semi-automatic segmentation. The semi-automatic method was based on a predefined sphere placed at the center of the tumor, as determined on CT. After initial segmentation, several thresholding steps (CT-based and PET-based) were applied to both methods to adjust for the relatively larger volume of the initial semi-automatic segmentation results. Quantitative parameters (including SUVmean, SUVmax, and MTV) were recorded and compared between the two methods, and the total segmentation time was also compared. Ergebnisse/Results: The semi-automatic method reduced segmentation time by 49% (mean difference -130,8 s; p<0.0001) and showed strong correlations with the conventional method across all recorded parameters (Spearman’s ρ≥0.67; all p<0.0001). Median paired biases ranged from+1.59 to+18.3 kBq/cc depending on applied thresholds, with statistically significant paired differences in the reported comparisons. Despite its reduced complexity, the semi-automatic method generated robust uptake recordings (including SUVmean and MTV), particularly evident in Spearman correlations. Schlussfolgerungen/Conclusion: The semi-automatic method reduced processing time by half while maintaining quantitative accuracy across all uptake parameters. It represents a practical and reproducible alternative to manual segmentation, well suited for high-throughput preclinical PET/CT studies. Publication History Article published online: 02 April 2026 © 2026. Thieme. All rights reserved. Georg Thieme Verlag KG Oswald-Hesse-Straße 50, 70469 Stuttgart, Germany
Authors 8
-
Affiliation as printed
Klinik für Nuklearmedizin, Uniklinik RWTH Aachen, Pauwelsstraße 30, 52074 Aachen, Deutschland
-
Affiliation as printed
Klinik für Nuklearmedizin, Uniklinik RWTH Aachen, Pauwelsstraße 30, 52074 Aachen, Deutschland
-
Maastricht University Medical Centre · Maastricht University · Universitätsklinikum Aachen
Affiliation as printed
Department of Imaging, Maastricht University Medical Center+, P. Debyelaan 25, 6229 HX Maastricht, Niederlande
Klinik für Nuklearmedizin, Uniklinik RWTH Aachen, Pauwelsstraße 30, 52074 Aachen, Deutschland
-
Affiliation as printed
Klinik für Nuklearmedizin, Uniklinik RWTH Aachen, Pauwelsstraße 30, 52074 Aachen, Deutschland
-
Affiliation as printed
Klinik für Nuklearmedizin, Uniklinik RWTH Aachen, Pauwelsstraße 30, 52074 Aachen, Deutschland
-
Affiliation as printed
Klinik für Nuklearmedizin, Uniklinik RWTH Aachen, Pauwelsstraße 30, 52074 Aachen, Deutschland
-
Maastricht University Medical Centre · Maastricht University · Universitätsklinikum Aachen
Affiliation as printed
Department of Imaging, Maastricht University Medical Center+, P. Debyelaan 25, 6229 HX Maastricht, Niederlande
Klinik für Nuklearmedizin, Uniklinik RWTH Aachen, Pauwelsstraße 30, 52074 Aachen, Deutschland
-
Maastricht University Medical Centre · Maastricht University · Universitätsklinikum Aachen
Affiliation as printed
Department of Imaging, Maastricht University Medical Center+, P. Debyelaan 25, 6229 HX Maastricht, Niederlande
Klinik für Nuklearmedizin, Uniklinik RWTH Aachen, Pauwelsstraße 30, 52074 Aachen, Deutschland
Cited by 0 stored of 0
No patents citing this paper on Lens.org (checked 2026-10-06).