TY - JOUR
T1 - Toward local arterial input functions in dynamic contrast-enhanced MRI.
AU - Fluckiger, Jacob U.
AU - Schabel, Matthias C.
AU - DiBella, Edward V.R.
PY - 2010/10
Y1 - 2010/10
N2 - To present a method for estimating the local arterial input function (AIF) within a dynamic contrast-enhanced MRI scan, based on the alternating minimization with model (AMM) method. This method clusters a subset of data into representative curves, which are then input to the AMM algorithm to return a parameterized AIF and pharmacokinetic parameters. Computer simulations are used to investigate the accuracy with which the AMM is able to estimate the true AIF as a function of the input tissue curves. Simulations show that a power law relates uncertainty in kinetic parameters and SNR and heterogeneity of the input. Kinetic parameters calculated with the measured AIF are significantly different from those calculated with either a global (P < 0.005) or a local input function (P = 0.0). The use of local AIFs instead of measured AIFs yield mean lesion-averaged parameter changes: K(trans): +24% [+15%, +70%], k(ep): +13% [-36%, +300%]. Globally estimated input functions yield mean lesion-averaged changes: K(trans): +9% [-38%, +65%], k(ep): +13% [-100%, +400%]. The observed improvement in fit quality with local AIFs was found to be significant when additional free parameters were accounted for using the Akaike information criterion. Local AIFs result in significantly different kinetic parameter values. The statistically significant improvement in fit quality suggests that changes in parameter estimates using local AIFs reflect differences in underlying tissue physiology.
AB - To present a method for estimating the local arterial input function (AIF) within a dynamic contrast-enhanced MRI scan, based on the alternating minimization with model (AMM) method. This method clusters a subset of data into representative curves, which are then input to the AMM algorithm to return a parameterized AIF and pharmacokinetic parameters. Computer simulations are used to investigate the accuracy with which the AMM is able to estimate the true AIF as a function of the input tissue curves. Simulations show that a power law relates uncertainty in kinetic parameters and SNR and heterogeneity of the input. Kinetic parameters calculated with the measured AIF are significantly different from those calculated with either a global (P < 0.005) or a local input function (P = 0.0). The use of local AIFs instead of measured AIFs yield mean lesion-averaged parameter changes: K(trans): +24% [+15%, +70%], k(ep): +13% [-36%, +300%]. Globally estimated input functions yield mean lesion-averaged changes: K(trans): +9% [-38%, +65%], k(ep): +13% [-100%, +400%]. The observed improvement in fit quality with local AIFs was found to be significant when additional free parameters were accounted for using the Akaike information criterion. Local AIFs result in significantly different kinetic parameter values. The statistically significant improvement in fit quality suggests that changes in parameter estimates using local AIFs reflect differences in underlying tissue physiology.
UR - http://www.scopus.com/inward/record.url?scp=79952202524&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=79952202524&partnerID=8YFLogxK
U2 - 10.1002/jmri.22339
DO - 10.1002/jmri.22339
M3 - Article
C2 - 20882623
AN - SCOPUS:79952202524
SN - 1053-1807
VL - 32
SP - 924
EP - 934
JO - Journal of magnetic resonance imaging : JMRI
JF - Journal of magnetic resonance imaging : JMRI
IS - 4
ER -