Speaker
Mr
Andreas Thon
(Philips Research Laboratories)
Description
Objectives: The determination of the exact shape of the input function is a
particular challenge in dynamic cardiac 82Rb PET perfusion studies. The dynamic
range of the activity in these studies easily covers two orders of magnitude.
Additionally, the injection bolus has a very sharp peak. The measurement of the
input function depends on both experimental errors (e.g. random noise, limited
spatial resolution) and data processing (e.g. frame durations). These issues are a
source of error in the assessment of the input function, especially regarding the
peak value and the decaying slope. Thus, the goal of our study is to address the
influence of such distortions of the input function on the model parameter K1,
which is linked to myocardial perfusion [1].
Methods: Using a one-compartment kinetic model [2,3,4], noise-free myocardium TACs
representing dynamic PET datasets were simulated with model parameters covering a
range of physiologic interest [3]. The TAC simulations are based on different input
functions computed with a generic analytical function. The parameters of this
function have been derived by numerical fits of various Rubidium PET measurements
of the left ventricular bloodpool.
The simulated input function and myocardium TACs have then been applied in a
kinetic analysis in the following way: the analytical input function has been
distorted by parameter variation, leading to different peak values and decaying
slopes. Then, the kinetic parameters have been re-estimated, using the respective
(unmodified) myocardium TAC as reference data. The (relative) bias of the K1
estimation was calculated and analyzed as a function of the input TAC variation,
and compared across the investigated range of kinetic model parameters.
Results: Underestimating (overestimating) the input peak value causes an
overestimation (underestimation) of K1, respectively. The magnitude of this effect
depends strongly on the blood volume fraction, and the FWHM of the input function.
This is because the wider the input peak, the stronger is the coupling of the input
function to the observed myocardium TAC. For reasonable values of the model
parameters, the relative bias in K1 is easily ±(10-30)% for a ±10% error in the
input peak. This causes an even larger bias in the blood flow values, due to its
nonlinear coupling to K1 [5].
Conclusion: Even with noise-free data, moderate errors in the estimation of the
input peak value lead to significant errors in the estimated K1 parameter.
Therefore, an accurate estimation of the input peak, e.g. by appropriate frame
durations, is necessary for a reliable kinetic analysis and blood flow estimation.
References:
[1] M. E. Phelps, “PET – Molecular Imaging and Its Biological Applications”,
Springer-Verlag, New York, 2004 (chapter 6)
[2] P.G. Coxson, R.H. Huesman, L. Borland, J. Nucl. Med. 1997; 38:660-667
[3] L. Golanowski, R. A. de Kemp, R. S. Beanlands, T. D. Ruddy, Proc. 22nd EMBS
Intern. Conf., July 23-28, Chicago IL, pp. 1096-1099
[4] G. El Fakhri et al, J. Nucl. Med. 2005; 46:1264-1271
[5] E. M. Renkin, Am. J. Physiol. 1959;197:1205-1210, and C. Crone, Acta Physiol.
Scand. 1964;58:292-305.
Author
Dr
Carsten Meyer
(Philips Research Laboraties)
Co-authors
Mr
Dragos-Nicolae Peligrad
(Philips Research Laboraties)
Mr
Martin Weibrecht
(Philips Research Laboraties)
