Nuklearmedizin 2020; 59(02): 173
DOI: 10.1055/s-0040-1708374
Wissenschaftliche Poster
Radiomics
© Georg Thieme Verlag KG Stuttgart · New York

A simple multi-modality phantom for simulating radiomic features in PET, CT and MRI

A Valladares
1   Medical University of vienna, QIMP Team, Center for Medical Physics and Biomedical Engineering, Vienna
,
T Beyer
1   Medical University of vienna, QIMP Team, Center for Medical Physics and Biomedical Engineering, Vienna
,
E Salomon
1   Medical University of vienna, QIMP Team, Center for Medical Physics and Biomedical Engineering, Vienna
,
I Rausch
1   Medical University of vienna, QIMP Team, Center for Medical Physics and Biomedical Engineering, Vienna
› Author Affiliations
Further Information

Publication History

Publication Date:
08 April 2020 (online)

 

Ziel/Aim Commonly-used phantom designs are not reflecting intra-tumoural heterogeneity in PET/CT and PET/MRI. We propose a simple concept for mimicking texture patterns in PET, CT and MR images and compare advanced heterogeneity metrics across imaging modality.

Methodik/Methods Three 50 mL-conical tubes were half-filled with different acrylic spheres sizes (S1: 1.6mm; S2: 50%(1.6mm)/50%(6.3mm), S3: 6.3 mm diameter) resulting in a homogeneous (spheres-free) and a heterogeneous area. Two replicates of these tubes were built and scanned (test-retest) in a PET/CT by using a standard oncological protocol, with 20 kBq/mL of [18F]FDG at the scan start. T1w- and T2w-MR images were also acquired in a 3T PET/MRI.

Mean CT and standardized uptake values (SUV), as well as 36 radiomic features, were derived from the homogeneous and heterogeneous regions of each tube. Coefficient of variation and reproducibility of textural features was assessed between replicates and for test-retest.

Ergebnisse/Results SUVmean in the homogeneous regions across all tubes and for test-retest was 4.48±0.04. SUVmean was for S1:1.70±0.01, S2:1.50±0.02 and S3:1.80±0.01. Mean CT values across the homogeneous regions were (11.1±0.6)HU. CT values for S1-S3 ranged between 79HU and 87HU.

High correlation between features values was found between replicates and between test-retest for all the modalities. In both cases, low variability (<5%) was obtained for 13, 16 and 30 features in PET, CT and MRI, respectively. Other individual features presented poor reproducibility, depending on the imaging modality and phantom sphere size. Here, variations up to 50%(S1), 20%(S2 and S3) for PET and CT, 20%(S1-S3) for T2w-MRI and 10%(S1-S3) for T1w-MRI were observed.

Schlussfolgerungen/Conclusions The similar variability of the features between test-retest and replicates indicates that the proposed phantoms can be used to provide means of features comparison between imaging systems. However, the feature variability seems highly feature and modality dependent; thus, the proposed phantom may only be useful for specific features.