TY - JOUR
T1 - “Stability of radiomic features in magnetic resonance imaging of the female pelvis: A multicentre phantom study” [Phys. Medica 130 (2025) 104895]
AU - Sghedoni, Roberto
AU - Origgi, Daniela
AU - Cucurachi, Noemi
AU - Minischetti, Giuseppe Castiglioni
AU - Alio, Davide
AU - Savini, Giovanni
AU - Botta, Francesca
AU - Marzi, Simona
AU - Aiello, Marco
AU - Rancati, Tiziana
AU - Cusumano, Davide
AU - Politi, Letterio Salvatore
AU - Didonna, Vittorio
AU - Massafra, Raffaella
AU - Petrillo, Antonella
AU - Esposito, Antonio
AU - Imparato, Sara
AU - Anemoni, Luca
AU - Bortolotto, Chandra
AU - Preda, Lorenzo
AU - Boldrini, Luca
PY - 2025
Y1 - 2025
N2 - The publisher regrets to miss the abstract of the above article. Please find the details below: Abstract Background. Radiomics has emerged as a promising tool for extracting large amounts of quantitative features from conventional biomedical imaging. Multicentre studies have investigated reproducibility and repeatability of radiomic features for different approaches in acquisition and image reconstruction for several imaging modalities. This study aimed to compare the impact of scanners and centre-specific parameter settings on the radiomic features calculated on images of a phantom. Methods. This is a MRI multicentre phantom study gathering ten centres and 19 scanners of the Italian “Alleanza contro il Cancro” network. The participant centres were asked to acquire repeated T2-weighted images of a dedicated pelvis-shaped phantom filled with silicon oil with four embedded inserts filled with polyethylene spheres and an agar solution, resulting in different heterogeneity textures. Each centre acquired the images using the local protocol for female pelvis investigation. For each MRI acquisition, radiomic features were extracted with and without image preprocessing. Results. Feature repeatability was assessed by calculating the coefficient of variation, while the agreement between different scanners was evaluated through the intraclass correlation coefficient. The coefficient of variation was lower than 7% for all the families of features and turned out to be unaffected by image preprocessing. Most families of features showed higher reproducibility when images underwent the preprocessing steps typically employed in an MR radiomic study. Discussion. Preliminary tests on phantoms can be preparatory to multicenter clinical radiomic studies to quantify the inherent variability in feature extraction due to the use of different MR scanners, sequences and image processing. The publisher would like to apologise for any inconvenience caused.
AB - The publisher regrets to miss the abstract of the above article. Please find the details below: Abstract Background. Radiomics has emerged as a promising tool for extracting large amounts of quantitative features from conventional biomedical imaging. Multicentre studies have investigated reproducibility and repeatability of radiomic features for different approaches in acquisition and image reconstruction for several imaging modalities. This study aimed to compare the impact of scanners and centre-specific parameter settings on the radiomic features calculated on images of a phantom. Methods. This is a MRI multicentre phantom study gathering ten centres and 19 scanners of the Italian “Alleanza contro il Cancro” network. The participant centres were asked to acquire repeated T2-weighted images of a dedicated pelvis-shaped phantom filled with silicon oil with four embedded inserts filled with polyethylene spheres and an agar solution, resulting in different heterogeneity textures. Each centre acquired the images using the local protocol for female pelvis investigation. For each MRI acquisition, radiomic features were extracted with and without image preprocessing. Results. Feature repeatability was assessed by calculating the coefficient of variation, while the agreement between different scanners was evaluated through the intraclass correlation coefficient. The coefficient of variation was lower than 7% for all the families of features and turned out to be unaffected by image preprocessing. Most families of features showed higher reproducibility when images underwent the preprocessing steps typically employed in an MR radiomic study. Discussion. Preliminary tests on phantoms can be preparatory to multicenter clinical radiomic studies to quantify the inherent variability in feature extraction due to the use of different MR scanners, sequences and image processing. The publisher would like to apologise for any inconvenience caused.
KW - Radiomic features
KW - Radiomic features
UR - https://publicatt.unicatt.it/handle/10807/313859
UR - https://www.scopus.com/inward/citedby.uri?partnerID=HzOxMe3b&scp=85218893467&origin=inward
UR - https://www.scopus.com/inward/record.uri?partnerID=HzOxMe3b&scp=85218893467&origin=inward
U2 - 10.1016/j.ejmp.2025.104948
DO - 10.1016/j.ejmp.2025.104948
M3 - Article
SN - 1120-1797
VL - 132
SP - N/A-N/A
JO - Physica Medica
JF - Physica Medica
IS - 132
ER -