Meeting Banner
Abstract #0240

Reproducibility of Dynamic Contrast-Enhanced MRI Perfusion Parameters on Various Computer Aided Diagnosis Workstations: Taking a Peek Into the Black Box.

MAGNA25Tobias Heye1, Matthew Davenport2, Jeff Horvath1, Sebastian Feuerlein1, Steven Breault1, Mustafa Bashir1, Elmar M. Merkle1, Daniel T. Boll1

1Department of Radiology, Duke University Medical Center, Durham, NC, United States; 2Department of Radiology, University of Michigan Health System, Ann Arbor, MI, United States

Although many factors contributing to overall measurement error have been identified, the effect of commercially available DCE-MRI post-processing solutions on quantitative (Ktrans, kep, ve) and semi-quantitative (iAUGC) pharmacokinetic parameters has yet to be defined. This study assessed the reproducibility of pharmacokinetic parameters between various commercially available post-processing solutions for DCE-MRI (Tissue4DTM, Siemens, Germany; DynaCADTM, Invivo, USA; AegisTM, Sentinelle Medical, Canada; CADvueTM; iCAD, Inc., USA). There is substantial variability (25.1-74.1% coefficient of variation) for DCE-MRI pharmacokinetic parameters across commercially available DCE-MRI post-processing solutions. If DCE-MRI is to succeed as a widely incorporated biomarker, the industry must agree on a post-processing standard.

Abstract PDF Presentation Video

Keywords

account accuracy aegis affect agree agreement aided alliance although angiography arterial assessment available background black bland boll calculations cancer clinical coefficient combination combinations commercially computer considered contrast contributing correlation currently davenport defined degrees delineated derivation derived diagnosis done duke dynamic efforts enable enhanced environment error establish female fibroid fibroids five function goodness health identified illustrates incorporated initial injection input interleaved intraclass introduced invasive lancet largest lesions limit limits magnitude many mapping material measured median medical model motion must output overall paradigm patients peek perfusion phys placement placing platforms post potential precision previously processed processing producing profile prospectively quantification quantitative radiology random randomly readers related reproducibility rescaled rescaling reside resolved revealed selected semi serve settings significantly since sites software solutions specified square standardized stochastic strong studies subject substantial succeed system table taking targeted tissue trajectories trans twist type underway unique upper user uterine variable variation various version volume widely workstation workstations years