- Poster presentation
- Open Access
System identification theory in myocardial perfusion modeling of dynamic contrast-enhanced MRI
© Vallée et al; licensee BioMed Central Ltd. 2009
- Published: 28 January 2009
- Myocardial Perfusion
- Myocardium Blood Flow
- Injection Rate
- Arterial Input Function
- System Identification Method
Tissue perfusion and blood volume can be quantified by MRI after contrast media (CM) injection and a one compartment model that yields K1, the blood to tissue transfer constant (related to myocardial blood flow MBF) and Vd, the CM distribution volume. Using system identification theory and computer simulations, the confounding effects of volume and rate of a contrast injection on the reliability of K1 estimation was recently evaluated (Aerts MRM 2008; 59:1111). The simulations predict that the K1 is most reliable with a high injection volume administered in a single injection, where high rates only modestly improve K1. However, the data used for the simulation were taken from a model of rectal cancer. It is not known how these simulations apply to myocardial perfusion at rest and at stress. Therefore, the aim of this study was to apply system identification theory on realistic data encountered in cardiac patients.
At rest, no significant difference in bias and standard deviation were observed with the identification process between the two shapes of the arterial input function (narrow and wide). Under-sampled curves introduce an increase of standard deviation of rest K1 and K2 but with no significant effect on the bias (< 1% for noise up to 5%). At stress (see Figure 1), there was no significant bias for both arterial input functions red and green curves of Figure 1) as well as for the wide undersampled function (black curve). However, almost 5× increase of bias was observed in the undersampled narrow function (blue curve). The standard deviation was increased by undersampling as well by a slow injection rate. Regarding the distribution volume Vd, accurate measurement was obtained for all the conditions (bias < 1% for a noise of 0.05 and standard deviation < 4%).
At rest, both slow and fast injection rates as well as undersampling up to 1 image/4 seconds yield accurate measurements of the myocardium blood flow and distribution volume. At stress the myocardium blood flow measurement is less accurate in case of a fast injection of a small volume of CM. The simulations indicated that MBF measurements at stress can improved by slowing the injection rate for the same sampling frequency.
This article is published under license to BioMed Central Ltd.