Premium
SU‐FF‐I‐110: Dynamic Model for Validation and Quality Assurance of Quantification of DCE Imaging
Author(s) -
Cao Y,
Li D,
Shen Z
Publication year - 2009
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.3181231
Subject(s) - quality assurance , noise (video) , dynamic contrast , software , computer science , approximation error , nuclear medicine , statistics , mathematics , medicine , artificial intelligence , radiology , pathology , magnetic resonance imaging , external quality assessment , image (mathematics) , programming language
: Dynamic contrast enhanced (DCE) MR imaging is emerging as a biomarker for assessment of cancer therapy and normal tissue toxicity. It is extremely important to standardize DCE MR imaging acquisition and evaluate software of pharmacokinetic (PK) quantification for multi‐center clinical trials. A synthesized dynamic model with known values of physiological parameters could provide a platform to evaluate PK modeling software. This study aims to create and validate a synthesized model for evaluation of quantification of the general Toft's model. Methods : The synthesized DCE model has K trans values from 0 to 0.5 min −1 , and ν p from 0.01 to 0.1, and k ep from 0.0 to 3.33 min −1 . The temporal sampling is 1 second. The dynamic series is ̃4 minutes. The Gaussian‐distributed random noise was added to the theoretic model to yield contrast‐to‐noise ratios (CNR) of 10, 20, 50, 100, and indefinite. For each parameter combination, the synthesized data were generated 2500 times for each of the CNR levels to evaluate the mean, relative SD, and relative error between the known and estimated values of the vascular parameters derived from quantification software. The relative SD and error assess the stability and accuracy (or bias) of an estimated parameter, respectively. Results and Discussions: Without noise, the estimated values of the three vascular parameters by our in‐house software are equal to the known values, suggesting the excellent quality of the dynamic model. With noise, the relative SDs and errors increase with a decrease in CNRs. For both stability and accuracy, ν p is superior compared to K trans and k ep , suggesting the different tolerances to noise. For K trans values of 0 and 0.001 and vp values of 0.1, stability and accuracy were substantial worse than other conditions, suggesting that it is necessary to determine acceptable parameter ranges with given CNRs and acquisition. Supported by 3P01CA59827