Dynamic contrast-enhanced MRI: Study of inter-software accuracy and reproducibility using simulated and clinical data

• Published in: Journal of magnetic resonance imaging 2016-06, Vol.43 (6), p.1288-1300
• Main Authors: Beuzit, Luc, Eliat, Pierre-Antoine, Brun, Vanessa, Ferré, Jean-Christophe, Gandon, Yves, Bannier, Elise, Saint-Jalmes, Hervé
• Format: Article
• Language: English
• Subjects: Adult; Aged; Bioengineering; Computer Simulation; Contrast Media - pharmacokinetics; DCE-MRI; Female; Humans; Image Interpretation, Computer-Assisted - methods; Imaging; inter-software variability; Kidney Neoplasms - diagnostic imaging; Kidney Neoplasms - metabolism; Kidney Neoplasms - pathology; Life Sciences; Magnetic resonance imaging; Magnetic Resonance Imaging - instrumentation; Magnetic Resonance Imaging - methods; Male; Meglumine - pharmacokinetics; Middle Aged; Models, Biological; NMR; Nuclear magnetic resonance; Organometallic Compounds - pharmacokinetics; Phantoms, Imaging; quantitative parameters; rectal cancer; Reproducibility of Results; Sensitivity and Specificity; simulated images; Software; Software Validation; Tofts model
• ISSN: 1053-1807; 1522-2586
• DOI: 10.1002/jmri.25101

Purpose To test the reproducibility and accuracy of pharmacokinetic parameter measurements on five analysis software packages (SPs) for dynamic contrast‐enhanced magnetic resonance imaging (DCE‐MRI), using simulated and clinical data. Materials and Methods This retrospective study was Institutional Review Board‐approved. Simulated tissues consisted of pixel clusters of calculated dynamic signal changes for combinations of Tofts model pharmacokinetic parameters (volume transfer constant [Ktrans], extravascular extracellular volume fraction [ve]), longitudinal relaxation time (T1). The clinical group comprised 27 patients treated for rectal cancer, with 36 3T DCE‐MR scans performed between November 2012 and February 2014, including dual‐flip‐angle T1 mapping and a dynamic postcontrast T1‐weighted, 3D spoiled gradient‐echo sequence. The clinical and simulated images were postprocessed with five SPs to measure Ktrans, ve, and the initial area under the gadolinium curve (iAUGC). Modified Bland–Altman analysis was conducted, intraclass correlation coefficients (ICCs) and within‐subject coefficients of variation were calculated. Results Thirty‐one examinations from 23 patients were of sufficient technical quality and postprocessed. Measurement errors were observed on the simulated data for all the pharmacokinetic parameters and SPs, with a bias ranging from −0.19 min−1 to 0.09 min−1 for Ktrans, −0.15 to 0.01 for ve, and −0.65 to 1.66 mmol.L−1.min for iAUGC. The ICC between SPs revealed moderate agreement for the simulated data (Ktrans: 0.50; ve: 0.67; iAUGC: 0.77) and very poor agreement for the clinical data (Ktrans: 0.10; ve: 0.16; iAUGC: 0.21). Conclusion Significant errors were found in the calculated DCE‐MRI pharmacokinetic parameters for the perfusion analysis SPs, resulting in poor inter‐software reproducibility. J. Magn. Reson. Imaging 2016;43:1288–1300.