1088 IEEE TRANSACTIONS ON MEDICAL IMAGING, VOL. 29, NO. 4, APRIL 2010
Correspondence
Comment on “Developing DCE-CT to Quantify
Intra-Tumor Heterogeneity in Breast Tumors
With Differing Angiogenic Phenotype”
Andrij Abramyuk, Gunter Wolf, Volker Hietschold,
Ulrike Haberland, Joerg van den Hoff, and Nasreddin Abolmaali
Abstract—In our comment some essential issues concerning determina-
tion of arterial input function (AIF), cardiac and respiratory related motion
artifacts, contrast agent application and compartmental model fitting done
by Cao et al., 2009 are discussed.
We have read with interest the paper by Cao et al. [1] illustrating
the applicability of dynamic contrast enhanced computed tomography
(DCE-CT) for the assessment of intratumoral heterogeneity in mouse
xenograft models in a clinical scanner. However, some major issues
concerning 1) determination of arterial input function (AIF), 2) car-
diac and respiratory related motion artifacts, 3) contrast agent appli-
cation, and 4) compartmental model fitting need to be discussed more
thoroughly. These issues may individually or collectively contribute to
the deviation of the results presented in Table II from literature. The
estimated values for fractional intravascular plasma and intersti-
tial volume in muscle are substantially different from those re-
ported under normal physiologic conditions ( : 0.5%–1.9% [2], [3];
: 13%–16% [3]–[5]). The fact that controversial values regarding
physiological parameters in normal tissue are given in literature raises
the question, to what extent reliable statements regarding such param-
eters could be applicable in tumor tissue.
1) A serious limitation of DCE-CT in small animals is defining an
appropriate vessel for determination of the AIF. If the vessel runs
obliquely through the imaging plane, the erroneously increased
standard deviation of the Gaussian curve would result in a lower
partial volume scoring factor and an underestimated AIF [6]. Ad-
ditionally, the analysis of the AIF may be influenced by beam
hardening artifacts, which mainly arise from heart and great ves-
sels induced by high intravascular iodine concentration through
rapid injection of large amounts of contrast material. The estima-
tion of different parameters of tumor vascularization may thus be
deteriorated by these artifacts, which might not easily be visible
Manuscript received June 25, 2009; revised August 25, 2009; accepted Au-
gust 29, 2009. First published October 02, 2009; current version published April
02, 2010.
A. Abramyuk is with the OncoRay, Medical Faculty Carl Gustav Carus,
Dresden University of Technology, 01307 Dresden, Germany (e-mail: Andrij.
Abramyuk@oncoray.de).
G. Wolf and N. Abolmaali are with the OncoRay, Medical Faculty Carl
Gustav Carus, Dresden University of Technology, 01307 Dresden, Germany
(e-mail: gunter.wolf@oncoray.de; nasreddin.abolmaali@oncoray.de).
V. Hietschold is with the Department of Diagnostic Radiology, University
Hospital Carl Gustav Carus Dresden, 01307 Dresden, Germany (e-mail: Volker.
Hietschold@uniklinikum-dresden.de).
U. Haberland is with the CT Division, Siemens Medical Solutions, 91301
Forchheim, Germany (e-mail: ulrike.haberland.ext@siemens.com).
J. van den Hoff is with the PET Center, Research Center Dresden-Rossendorf,
01314 Dresden, Germany (e-mail: j.van_den_hoff@fzd.de).
Digital Object Identifier 10.1109/TMI.2009.2031780
depending on window settings but are strongly expected according
to the positioning of mice in the scanner.
Despite the mouse small heart size with an enddiastolic left ven-
tricular diameter of 3.2 mm [7] and a contraction rate of more than
600 beats per minute [8], the application of the 90th percentile
value derived from a region of interest (ROI) in the left ventricle
seems to be a reasonable approach for the determination of the
AIF. Nevertheless, motion influence (both from heart contraction
and breathing) as well as partial volume effects will be present and
the question of systematic and/or statistical errors in deriving the
AIF should be addressed.
2) Furthermore, respiratory motion artifacts and partial volume ef-
fects are also relevant for ROI and pixel based evaluation in small
tumors (mean diameter of tumors used in this study was approxi-
mately 6 mm). This is even more relevant after further tumor divi-
sion into up to eight subvolumes. Respiration will cause relevant
translation and most probably rotation of tumors injected into the
mammary fat pad. Therefore, it remains unclear how influences
of thorax cage movement due to breathing were avoided. Did the
authors use the 90th percentile value derived from a ROI for res-
piratory motion correction (as was done for AIF) in tumor eval-
uations as well? This strategy optimizes for the correct coverage
of high concentrations in a surrounding of low concentrations. If
applied in the tumor—would this strategy be adequate for inho-
mogeneous tumor masses?
3) An essential prerequisite for the estimation of different kinetic pa-
rameters from the quantification of the time course of a tracer or
contrast agent in blood pool or tissue is, that any tracer should
be applied in concentrations that do not perturb the hemodynamic
status of the investigated system [9], [10]. In general, the appli-
cable volume of administered substances that are recommended
for intravenous bolus injection (carried out over a relatively short
period of time of approximately 1 min) in mice should not ex-
ceed 5 ml/kg body weight [8], [11], [12]. Thus, for mice of 25 g
body weight the intravenous bolus injection volume should to-
tally not exceed 125 , if the physiological conditions should
not be disturbed significantly. However, in their study the au-
thors intravenously injected a bolus of 200 of contrast agent,
which exceeds the recommended volume considerably (by 60%).
It has to be kept in mind that parameter estimations derived from
such experimental conditions would not necessarily allow for con-
clusions concerning the (patho-) physiological conditions of the
tissue under investigation. In addition, due to the large volume
administered it can not be excluded that further aspects like the
elasticity of the vessel walls influence the estimation of perfusion
parameters.
4) Regarding the used two-compartment model and the applied
quantification procedure a few remarks seem necessary.
a) The model contains four free parameters which are deter-
mined simultaneously. Usefulness of this model depends
critically on the ability to differentiate the actual impulse re-
sponse of the capillary space from an instantaneous (“delta”)
response to the given input. We think the authors should have
commented more clearly on the practical limitations of using
such a model, specifically the need for very high systematic
and statistical accuracy of the dynamic data combined with
very high temporal resolution and the critical importance
0278-0062/$26.00 © 2010 IEEE