The extraction of radiomic features from CT images is an analysis tool that allows the identification of quantitative information not perceptible to the human eye. Unfortunately, its potential is limited by the dependence of the features on the image acquisition parameters. Doing an upstream harmonization based on image quality, we characterized the features extracted from the RadiomiK phantom, an in-house developed geometric phantom containing inserts with different materials and textures. Thirty CT images of RadiomiK were acquired with ten protocols using two scanners, three dose values and four levels of iterative reconstruction algorithms. The repeatability on three repetitions of the same protocol and on ROIs placements and volumes was assessed by calculating the coefficient of variation; the robustness by computing the intraclass correlation coefficient among different protocols. The harmonization improved the robustness of the features, especially for more homogeneous inserts. A standardized feature extraction process should be used as the ROI placement and volume strongly influenced the repeatability.
Radiomic features characterization of CT images using the RadiomiK phantom.
Tenerani M. I.
Primo
;Imbriani M.;Lizzi F.;Quattrocchi M.;Scapicchio C.;Zafaranchi A.;Fantacci M. E.Ultimo
2024-01-01
Abstract
The extraction of radiomic features from CT images is an analysis tool that allows the identification of quantitative information not perceptible to the human eye. Unfortunately, its potential is limited by the dependence of the features on the image acquisition parameters. Doing an upstream harmonization based on image quality, we characterized the features extracted from the RadiomiK phantom, an in-house developed geometric phantom containing inserts with different materials and textures. Thirty CT images of RadiomiK were acquired with ten protocols using two scanners, three dose values and four levels of iterative reconstruction algorithms. The repeatability on three repetitions of the same protocol and on ROIs placements and volumes was assessed by calculating the coefficient of variation; the robustness by computing the intraclass correlation coefficient among different protocols. The harmonization improved the robustness of the features, especially for more homogeneous inserts. A standardized feature extraction process should be used as the ROI placement and volume strongly influenced the repeatability.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.