Magnetic Resonance in Medicine 56:340 –347 (2006)
In Vivo Quantification of Regional Myocardial Blood Flow: Validity of the Fast-Exchange Approximation for Intravascular T1 Contrast Agent and Long Inversion Time Marlene Wiart,1 Sabin Carme,1,2 Wilfried Maı¨,1 Henrik B.W. Larsson,3,4 Bruno Neyran,1 and Emmanuelle Canet-Soulas1* In the present study we investigated the effects of water exchange between intra- and extravascular compartments on absolute quantification of regional myocardial blood flow (rMBF) using a saturation-recovery sequence with a rather long inversion time (TI, 176 ms) and a T1-shortening intravascular contrast agent (CMD-A2-Gd-DOTA). Data were acquired in normal and ischemically injured pigs, with radiolabeled microsphere flow measurements used as the gold standard. Five water exchange rates (fast, 6 Hz, 3 Hz, 1 Hz, and no exchange) were tested. The results demonstrate that the fast-exchange approximation may be appropriate for rMBF quantification using the described experimental setting. Relaxation rate change (⌬R1) analysis improved the accuracy of the analysis of rMBF compared to the MR signal. In conclusion, the current protocol could provide sufficient accuracy for estimating rMBF assuming fast exchange and a linear relationship between signal and tissue concentration when quantification of precontrast T1 is not an option. Magn Reson Med 56:340 –347, 2006. © 2006 Wiley-Liss, Inc. Key words: magnetic resonance imaging; contrast medium; water exchange; regional blood flow; myocardium ischemia.
First-pass dynamic contrast-enhanced (DCE)-MRI using T1-weighted imaging and an intravascular contrast agent may be a promising tool for the functional assessment of coronary artery stenosis (1,2). The principles of indicatordilution theory can be applied to quantify regional myocardial blood flow (rMBF) provided that MR-signal variations can be related to contrast agent concentrations. The most direct approach to the measurement of concentration is to quantify the change in T1 relaxation rate, ⌬R1, itself (3,4). Although the contrast agent may be confined within the blood compartment, its passage through the capillaries may still have an impact on the T1 of the extravascular space if a water proton that has “sensed” the paramagnetic
1
Creatis UMR CNRS 5515 U630 INSERM, Lyon, France. Guerbet, Aulnay-sous-Bois, France. 3 Functional MR Unit, Glostrup Hospital, Copenhagen University, Copenhagen, Denmark. 4 Department of Circulation and Medical Imaging, Faculty of Medicine, Norwegian University of Science and Technology, Trondheim, Norway. Grant sponsor: French Ministry of Research and Technology; Grant number: 00-2-11922; Grant sponsor: Rhoˆne Alpes regional authority; Grant number: 0100898801; Grant sponsors: Guerbet; Claude Bernard Lyon I University. Presented in part at the 12th Annual Meeting of ISMRM, Kyoto, Japan, 2004. *Correspondence to: Emmanuelle Canet-Soulas, Creatis UMR CNRS 5515 U630 Inserm, Hoˆpital Neurocardiologique, 28 Avenue du Doyen Lepine, B.P. Lyon-Montchat, 69394 Lyon Cedex 03, France. E-mail:
[email protected] Received 19 October 2005; revised 16 March 2006; accepted 17 April 2006. DOI 10.1002/mrm.20969 Published online 6 July 2006 in Wiley InterScience (www.interscience.wiley. com). 2
© 2006 Wiley-Liss, Inc.
molecules diffuses into the extravascular space before it is measured. The influence of water exchange between the intra- and extravascular spaces on the MR signal is well known (5–9) and has been widely addressed in the myocardium (10 –16). Optimum pulse sequence parameters have been proposed to minimize the dependence of MRI contrast enhancement on water exchange (7,10,13,16,17). For instance, a short inversion time (TI ⬍ 100 ms) will leave little time for water exchange to influence magnetization before measurement, and hence signal changes will be approximately exchange-independent (14). However, a short TI also considerably reduces the dynamic range of the first-pass experiment. In a previous study we tested a prototypic blood-pool contrast agent, CMD-A2-Gd-DOTA (P717; Guerbet, Aulnay-sous-Bois, France) for in vivo quantification of rMBF and the myocardial perfusion reserve (MPR) (18). The imaging protocol, which was designed to be as close as possible to the clinical setting, in which hypoperfused myocardial territories are diagnosed by visual inspection of the image series, involved a rather long TI (176 ms) to enhance contrast (18). Questions then arose regarding the possible confounding effects of water exchange on the quantification of perfusion parameters from these data. The first aim of the present study was to investigate water exchange effects on absolute rMBF and MPR quantification in a clinical-like context, in normal and in ischemically-injured pigs, by comparison with radiolabeled microsphere flow measurements. Five exchange rates (from fast exchange to no exchange) were explored, with intermediate values of 6, 3, and 1 Hz chosen in accordance with previous myocardium studies (7,11,15). Given a sufficiently fast exchange of water between the intra- and extravascular compartments, a linear relationship will exist between the tissue contrast agent concentration and the tissue change in R1. However, ⌬R1 could have a linear relation to concentration and not to the MR signal (11). Our second aim was to evaluate the contribution of ⌬R1 analysis compared to signal analysis using the fast-exchange approximation. THEORY The central equation for estimating rMBF (in ml/min/g of tissue) using an intravascular indicator is given by: C t共t兲 ⫽ 䡠 rMBF 䡠 C a共t兲 䊟 R共t兲
[1]
where V denotes the convolution product, Ct(t) is the tissue concentration of the indicator (in mM), Ca(t), the
340
Myocardial Water Exchange and Blood Flow
341
FIG. 1. Conversion of signal-time curve to concentration-time curves according to different water-exchange rates. a: T1-weighted first-pass images of the myocardium of pig #4 (contoured in black): 1) precontrast image, 2) arrival of the contrast agent in the right ventricle, 3) arrival of the contrast agent in the left ventricle, 4 and 5) passage of the contrast agent through the myocardium, and 6) wash-out. b: Average signal-time curve in the myocardium in arbitrary units (a.u.). Images 1– 6 indicate the steps described in part a. c: MR signal as a function of ⌬R1 for different water exchange rates, from fast exchange (Fast) to no exchange (No) through exchange rates corresponding to realistic values for the myocardium (1/ ⫽ 6, 3, and 1 Hz). d: The concentration-time curve in the myocardium can be constructed for each time-point for the five hypotheses.
concentration of the indicator in the arterial input function (AIF, in mM), is the tissue density (in g/ml), and R(t) is the residue function of the system (i.e., the fraction of indicator still present in the tissue at time t if the AIF was given at time zero as an ideal instantaneous bolus of amplitude 1 (no dimension)). For an intravascular contrast agent, tissue concentration is equal to capillary blood concentration, Cb(t), weighted by the fractional blood volume, fBV (in %): C t共t兲 ⫽ fBV 䡠 C b共t兲
[2]
According to both nuclear relaxation theory and in vitro measurement (19), the change in relaxation rate induced by the contrast agent in the blood, ⌬R1b(t), is proportional to the contrast agent concentration at low dose: C b共t兲 ⫽
冉
冊
1 1 1 䡠 ⫺ r 1 T 1post共t兲 T 1pre
⫽ b
1 䡠 ⌬R 1b共t兲 r1
[3]
where T1pre is the precontrast T1, T1post(t) is the postcontrast T1 at each time t, and R1 is the longitudinal relaxivity of the contrast agent. This relationship can be translated for the AIF straightforwardly: 1 C a共t兲 ⫽ 䡠 ⌬R 1a共t兲 r1
[4]
1 䡠 ⌬R 1b共t兲 r1
1 1 1 ⫽ ⫹ b e
[5]
T1pre can be estimated using a dedicated sequence, such as an inversion-recovery prepared sequence. T1post(t) should be deduced from the MR signal measured in the first-pass images. Conversion between the MR signal and T1post depends on the hypothesis made as to the water exchange
[6]
where b and e are the average contrast agent residence times, respectively, in the intravascular compartment (blood, i.e., plasma plus erythrocytes, assuming fast exchange between these) and extravascular compartment (i.e., interstitium plus intracellular space, assuming fast exchange between these as well). While the exchange rate itself is constant, its characterization as fast, intermediate, or slow depends on the difference in relaxation rate between the two spaces. Figure 1 shows an example of conversion of MR signal to tissue concentration for the water exchange rates detailed below. Let T1b denote the longitudinal relaxation time in the blood, and T1e the longitudinal relaxation time in the extravascular space.
Fast Exchange
冏
1 1 1 ⫺ ⬎⬎ T1b T1e
冏
The tissue relaxes in a single relaxation rate, 1/T1t, equal to the sum of the relaxation rates of the compartments weighted by their respective fractional volumes (7): 1 fBV 1 ⫺ fBV ⫽ ⫹ T 1t T 1b T 1e
For the tissue, combining Eqs. [2] and [3] gives: C t共t兲 ⫽ fBV 䡠
rate. Let us define the net exchange rate, 1/, between the intra- and extravascular spaces:
[7]
Assuming the same initial water T1’s in the blood and extravascular compartment (7), it can be shown that ⌬R 1t共t兲 ⫽ fBV 䡠 ⌬R 1b共t兲 Thus, Eq. [5] can be written as
[8]
342
Wiart et al.
C t共t兲 ⫽
1 䡠 ⌬R 1t共t兲 r1
[9]
For the fast-exchange approximation, the following TurboFLASH sequence equation was fitted to the measured signal S at each time t, by nonlinear least-squares minimization, using T1 as the model parameter (20):
冏
冋
TI
case again, Eqs. [8] and [9] do not hold true. The measured MR signal S can be described as the sum of signals from the two compartments, weighted by their respective fractional blood volumes (11): S ⫽ fBV 䡠 S共T 1b兲 ⫹ 共1 ⫺ fBV兲 䡠 S共T 1e兲 which gives:
TR
S共T 1兲 ⫽ ⍀ M 0 共1 ⫺ e ⫺T1兲 䡠 cos n⫺1␣ 䡠 e ⫺共n⫺1兲 T1
S共T 1b兲 ⫽
册冏
[13]
1 䡠 关S ⫺ 共1 ⫺ fBV兲 䡠 S共T 1e兲兴 fBV
[14]
TR
⫹ 共1 ⫺ e
TR T1
⫺
兲
1 ⫺ cos n⫺1␣ 䡠 e ⫺共n⫺1兲 T1 1 ⫺ cos␣ 䡠 e
TR ⫺ T1
[10]
where ⍀ is the receiver gain, M0 is the fully relaxed magnetization, ␣ is the flip angle, and n ⫽ Nphase/2 is the number of phase-encoding steps to reach ky ⫽ 0. Equation [4] was used to assess concentration in the AIF, since exchange is considered fast in the blood (7). Equation [9] was used to assess tissue concentration when the fastexchange approximation was assessed.
Intermediate Exchange
冏
1 1 1 ⫺ ⬇ T1b T1e
冏
[11]
[12]
Note that in this case an assumption has to be made for the value of fBV. Equation [5] was then used to evaluate contrast agent concentration in the tissue.
冏
1 1 1 ⫺ ⬍⬍ T1b T1e
冋冉
TI 䡠 cos n⫺1␣ ⫹ TR 䡠
冊 册
1 ⫺ cos n⫺1␣ 1 1 ⫺ cos␣ T 1
[15]
hence: S共T 1兲 ⫽ c 䡠
1 T1
C t/a共t兲 ⫽ K 䡠 共S共t兲 ⫺ S共0兲兲
where b denotes the vascular space (blood), e is the extravascular space, M(t) is the magnetization at time t, and M0 is the fully relaxed magnetization. Solutions to these coupled equations are given in Ref. 11. The solutions were iteratively fitted to the measured signal S at each time t, using T1b and T1e as model parameters with the assumption (21):
No Exchange
Provided that (n – 1) 䡠 TR ⬍⬍ T1, and TI ⬍⬍ T1, Eq. [10] can be approximated by:
[16]
Using the fast-exchange approximation experiment (T1(t)), Eqs. [4] and [9] are then equivalent to:
dM b共t兲 M 0b ⫺ M b共t兲 M b共t兲 M e共t兲 ⫺ ⫹ , ⫽ dt T 1b b e
e 1 ⫺ fBV ⫽ b fBV
Linearity Approximation Between the MR Signal and R1
S共T 1兲 ⫽ ⍀ 䡠 M 0
When the water exchange is slower, the intra- and extravascular spaces relax with separate time constants, so that Eq. [8], and therefore Eq. [9], are no longer valid. The description of the resulting magnetization is complex, but a two-site exchange model can be used, adding a term accounting for water exchanges to the Bloch equations (11):
dM e共t兲 M 0e ⫺ M e共t兲 M e共t兲 M b共t兲 ⫺ ⫹ ⫽ dt T 1e e b
S(T1e) can be approximated with the help of Eq. [10] using the measured precontrast T1 in tissue. In the no-exchange case, an assumption also has to be made for the value of fBV. Equation [10] was fitted to the calculated signal S(T1b) by nonlinear least-squares minimization, using T1b as the model parameter. Equation [5] was finally used to evaluate contrast agent tissue concentration.
冏
If no water exchange occurs during the experiment, the tissue will relax with two distinct relaxation rates. In this
[17]
If Eq. [17] is valid (and the fast-exchange approximation acceptable), there is no need to quantify precontrast T1. The value of the proportionality constant K is assumed to be the same for the artery and the tissue, and therefore it simplifies in Eq. [1]. MATERIALS AND METHODS Experimental Protocol The study conformed to National Institutes of Health guidelines for the care and use of laboratory animals. rMBF was assessed using radioactive microspheres and DCE-MRI in four control pigs and three pigs with an experimental coronary stenosis (a total of seven farm pigs, weighing 30 kg), at rest and under pharmacological stress induced by 140 g/kg/min adenosine infusion (Sigma, Lyon, France) for 10 min. Animal Preparation and Surgery The animals were sedated with an intramuscular cocktail of droperidol (0.8 mg/kg; Janssen-Cylag, Issy-les-Moulineaux, France), ketamine (15 mg/kg; Panpharma, Fouge`res,
Myocardial Water Exchange and Blood Flow
343
France), atropine (0.04 mg/kg; Aguettant, Lyon, France), and butorphanol (4 mg/kg; Fort Dodge, IA), and anesthesia was induced by administration of propofol through a catheter placed in the marginal vein of the ear (4 mg/kg; Zeneca Pharma, Cergy, France). Anesthesia was maintained during the experiment by continuous infusion of propofol at a rate of 400 mg/hr. After the heart was exposed by left lateral thoracotomy, 6 mg of pancuronium bromide (Organon Teknika, Fresnes, France) were administered to suppress respiratory movement during scanning, and the pigs were artificially ventilated by a Siemens-Elena 900 ventilator. A ventriculography catheter was inserted in the left atrium for microsphere administration. A catheter was placed in the carotid artery to withdraw blood samples and to monitor systemic blood pressure. Another catheter was positioned in the jugular vein for contrast agent and adenosine administration. Significant coronary stenosis was produced by dissecting the left circumflex coronary artery free from surrounding tissue, placing a transit-time flow probe connected to a digital flowmeter (Transonic®; EMKA Technologie, Paris, France) around the vessel, and inflating a balloon occluder.
agent blood concentration. The same MR protocol was repeated 10 min after induction of pharmacological stress.
Contrast Agent
Image and Data Analysis
CMD-A2-Gd-DOTA (P717; Guerbet, Aulnay-sous-Bois, France) is a carboxymethyl-dextran polymer substituted with the paramagnetic macrocyclic complex Gd-DOTA. It has an average molecular weight of 50.5 kDa and presents the overall characteristics of a blood-pool contrast agent (1). The longitudinal relaxivity is high compared to conventional low-molecular-weight gadolinium chelates (R1 ⫽ 9.4 mM–1.s–1 at 60 MHz in water at 37°C). The contrast agent was administered at a dose of 0.009 mmol Gd/kg.
All MR image data were transferred to a personal computer (1.6 GHz, 256 Mbytes) and analyzed using in-house-developed software written in Matlab script (MathWorks, Natick, MA, USA). Left ventricle endocardial and epicardial borders were drawn manually for both sequences. Precontrast T1 values were estimated by nonlinear fitting of the inversion-recovery multi-TI data. An AIF was selected in the cavity of the left ventricle for each dynamic series. The signal-time curve representing the AIF was converted into a concentration-time curve assuming fast water exchange (7). Gadolinium blood sample concentration was determined by inductively coupled plasma mass spectroscopy (ICP-MS). MR-derived and actual contrast agent blood concentration-time curves were co-recorded in time and compared to test the linearity assumption of Eq. [4]. Average signal-time curves were generated for the whole myocardium in control pigs (one myocardial region of interest (ROI) per slice). In injured pigs, average signaltime curves were generated in sectors subjected to ischemia according to the ex vivo microsphere blood flow segmentation and in sectors that had not been subjected to ischemia (two ROIs per slice). Each signal-time curve was converted into five concentration-time curves Ctissue(t) according to different water exchange hypotheses: 1) fast exchange, 2) 1/ ⫽ 6 Hz, 3) 1/ ⫽ 3 Hz, 4) 1/ ⫽ 1 Hz, and 5) no exchange. Fractional blood volume was assumed to be 10% at rest and 15% under stress (7). Deconvolution of Eq. [1] was performed with ⫽ 1 g/ml using the ARMA model, as previously described (23). This approach allows one to estimate two perfusion parameters (the rMBF and regional myocardial blood volume (rMBV)) independently based on first-pass data (23). For a given ROI, six MR-derived rMBF values were computed—first assuming Eq. [17] (i.e., deconvoluting the intensity-time
MRI MRI was performed immediately after the thorax was closed, on a 1.5T MR scanner (Vision, Siemens®, Erlangen, Germany) with the animal in right decubitus and a four-channel phased-array coil wrapped around the chest wall. All images were acquired using T1-weighted TurboFLASH sequences with the following parameters: TR/ TE ⫽ 2.4 ms/1.2 ms; flip angle ␣ ⫽ 8°; field of view (FOV) ⫽ 188 ⫻ 300 mm; matrix ⫽ 256 ⫻ 256; slice thickness ⫽ 8 mm; and two short-axis slices (basal and midventricular). A set of 10 inversion-recovery prepared images with TI varying between 20 and 3000 ms was first acquired for precontrast longitudinal relaxation rate quantification. The ventilator was then switched off for dynamic “breath-hold” imaging. First-pass images were acquired before, during, and after bolus injection of the contrast agent with a saturation-recovery prepared sequence (TI ⫽ 176 ms, number of phases Nphase ⫽ 112). Sixty images were acquired per slice (for a total of 120 images per examination) at a rate of two slices every two heartbeats using ECG gating, resulting in a temporal resolution of 2 s. Blood samples were taken from the carotid every 2 s during first-pass imaging to measure the actual contrast
Postmortem Radioactive Microsphere Analysis Approximately 106 radioactive microspheres (diameter ⫽ 11.4 ⫾ 0.1 m; activity ⫽ 0.06 – 0.1 mCi) were injected into the left atrium immediately before each dynamic MR sequence (141Ce for the rest condition, and 113Sn for the stress condition; NEN-TRAC®, Life Science Products Inc., Boston, MA, USA). For calibration, a reference withdrawal was started 30 s before injection with a power injector (Perfuser®, Braun, Germany) at 10 ml/min and continued for 1 min. After completion of imaging the animals were euthanized with an overdose of pentobarbital, and the heart was retrieved for microsphere analysis. Two slices matching the MRI image contour were taken and divided into 12 sectors according to AHA standards (22). rMBF (in ml/ min/100 g) was estimated from radioactivity measurements performed in each tissue sample with a singlephoton counter (Cobra II Auto-Gamma Packard®; Packard Instruments Company, Meriden, CT, USA), as described elsewhere (1).
344
Wiart et al.
FIG. 2. MR-derived AIF represented as ⌬R1 curve (gray, diamonds) and corresponding curve generated from gadolinium concentrationtime curves in blood samples (black, square) for pig #4.
curves obtained by subtracting the signal-time curve from the mean precontrast signal), and second on the basis of the contrast agent concentration curves obtained on the five tissue water exchange hypotheses using the ⌬R1 analysis. Also, six myocardial perfusion reserve (MPR) values were estimated for each ROI as the ratio of rMBF during hyperemia divided by rMBF at rest. Values of rMBV obtained using R1 analysis under the fast-exchange hypothesis were analyzed in control pigs at rest and under stress to assess the validity of our fBV value assumption. Values obtained in ROIs from the two slices were averaged to reduce inaccuracy resulting from coregistration between the microsphere data and the in vivo images (14). Statistical Analysis All values are expressed as the mean ⫾ standard deviation (SD). Bland-Altman plots and scatterplots were used to compare the methods (24). RESULTS Carotid sample blood concentration peaks were in the 0.4 – 0.6 mM range. MR-derived AIF curves obtained with the fast-exchange approximation were in good agreement with the reference concentration curves in the carotid (Fig. 2), with left ventricle ⌬R1 peaks in the 3– 8 s–1 range (calculated concentration ⫽ 0.3– 0.8 mM). In control tissue at rest, ⌬R1 was one order of magnitude lower than in blood (range ⫽ 0.2–1.2 s–1; Fig. 1d). At rest the rMBV was 9.0% ⫾ 0.5%, and under stress it was 16.0% ⫾ 3.0%. Impact of Water Exchange on rMBF and MPR Quantification Figure 3 shows rMBF values in ischemic and control ROIs at rest and under stress, obtained with microspheres on the one hand and with MRI for each water-exchange hypothesis on the other hand. In this experimental setting the fast-exchange approximation best approximated microsphere measurement of myocardial perfusion. MPR values obtained in ischemic and control pigs are shown in Table 1. The fast-exchange hypothesis underestimated MPR in control pigs compared
FIG. 3. Mean rMBF obtained with microspheres and MRI using the five different assumptions for water exchange at rest and under adenosine-induced stress in control (N ⫽ 4) and ischemic (N ⫽ 3) pigs.
to the intermediate-exchange regime. In ischemically injured pigs, MPR values were more dispersed with the fast-exchange approximation than with microspheres, but the average values were identical. Impact of Signal Analysis vs. ⌬R1 Analysis on rMBF and MPR Quantification Both signal and T1 analyses showed an increase of the mean differences and SD of the differences with flow, and therefore bias and 95% limit intervals were calculated as a function of mean flow (Fig. 4a). Bias slope and intercept were higher for signal analysis than for T1 analysis with fast exchange (bias slope ⫽ 0.65 vs. 0.33; Fig. 4a). However, the confidence interval was narrower using signal analysis (Fig. 4a). rMBF was overestimated compared to microsphere blood flow by both methods, but less so on ⌬R1 analysis (slope of linear regression line ⫽ 1.5 vs. 1.1; Fig. 4b). Other approximations than fast exchange resulted in higher bias slope and further overestimation of rMBF (data not shown): ⫽ 6 Hz, bias slope: 0.72; ⫽ 3 Hz, 0.86; ⫽ 1 Hz, 1.06; no exchange, 1.17. MPR was underestimated in control and ischemic pigs by signal analysis (Table 1).
Table 1 Myocardial Perfusion Reserve (MPR) Measured in % as the Ratio rMBF (Stress):rMBF (Rest)
Microspheres MRI Signal analysis Fast exchange MRI T1 analysis Fast exchange 1/ ⫽ 6 Hz 1/ ⫽ 3 Hz 1/ ⫽ 1 Hz No exchange
Control pigs (N ⫽ 4)
Ischemic pigs (N ⫽ 3)
4.0 ⫾ 1.8
1.4 ⫾ 0.8
3.0 ⫾ 1.1
1.0 ⫾ 0.6
3.0 ⫾ 1.0 3.7 ⫾ 1.5 3.9 ⫾ 1.6 4.3 ⫾ 1.7 4.4 ⫾ 1.6
1.4 ⫾ 1.4 1.5 ⫾ 1.5 1.3 ⫾ 1.1 1.3 ⫾ 1.1 1.5 ⫾ 1.7
Myocardial Water Exchange and Blood Flow
345
FIG. 4. a: Bland-Altman plots comparing microsphere- and MRI-derived rMBF for signal analysis (diamond, black) and for R1 analysis (square, gray). Dashed lines represent bias and plain lines represent the 95% limit interval. b: Scatterplot of MBF measured using the microsphere- and MRI-derived rMBF for signal analysis (diamond, black) and R1 analysis (square, gray). Both methods assume fast exchange between the intra- and extravascular spaces, and deconvolution was performed using the ARMA model with rMBF in ml/min/ 100 g.
DISCUSSION This study shows that the best estimates of rMBF compared to microsphere measurements were obtained with the fast-exchange approximation in the present specific clinical-like protocol. Relaxation rate change (⌬R1) analysis improved rMBF quantification compared to MR-signal analysis. Impact of Water Exchange on rMBF and MPR Quantification For the dose and TI used in the present study, the relationship between MR-signal and tissue concentration was linear under the fast-exchange approximation and nonlinear in the other cases (Fig. 1c). As predicted (Fig. 1d), rMBF estimates increased as the exchange became slower (Fig. 3). This increase in rMBF was more marked under stress than at rest. This was not unexpected, because the gap between the curves representing the signal as a function of concentration increase as the signal increases (Fig. 1c), and the signal peak is higher under stress than at rest. Reciprocally, water exchange did not impact rMBF values much in ischemic regions, where signal peaks remained low. The closest estimates of blood flow were obtained using the fast-exchange approximation, which could be explained by exchange rates higher than 6 Hz. Bjornerud et al. (16) reported net exchange rate values of 4 –23 Hz for myocardium in eight pigs. The concept of perfusion reserve was developed to determine the functional significance of a coronary artery stenosis (25). In the injured pigs in the present study the overall perfusion reserve was reduced, in agreement with previous studies (26 –28). The ⌬R1 approach with the fast-
exchange approximation underestimated the MPR in controls, which may be due to the administration of a second dose of contrast agent (stress) approximately 60 min after the first injection (rest). Since the contrast agent used in this study had a long vascular remanence, peak contrast agent concentration could have reached the limit of linearity during the stress experiment, resulting in underestimation of rMBF and hence of MPR. The signal-to-noise ratio (SNR) was deteriorated in ischemic regions, where blood flow was very low, which could explain the rather high MPR SD in these subjects. Some sources of error in the present study need to be considered. First, an assumption had to be made about the fBV for the intermediate- and no-exchange models. The fBV values selected in the light of the literature (10% at rest and 15% under stress (7)) matched the estimates of regional myocardial blood volume in the control pigs in the study when ⌬R1 analysis was applied with the fastexchange approximation (9.0% ⫾ 0.5% at rest, and 16% ⫾ 3% under stress). It is well known that changes in the fractional size of the compartments affect the magnitude of the change in signal, particularly at long TI values (14). The assumption regarding fBV during pharmacological stress might break down, especially in ischemic pigs, since vasodilatation may be altered in these subjects. This would lead to overestimating MRI-derived MPR in ischemic pigs compared to the microsphere results. Second, the exchange rate was considered to be unchanging from rest to stress, for purposes of MPR assessment. The water exchange rate is proportional to the permeability surface (PS) area product and inversely proportional to the intravascular volume, V. Under stress, both PS and V will increase and, assuming a cylindrical shape
346
Wiart et al.
for the blood vessel as a first approximation, PS/V and thus 1/ will decrease. Water exchange should therefore be slower under stress than at rest. Still assuming a cylindrical vessel, the fBV increase (from 10% to 15%) assumed in the present study resulted in a 20% diminution of the exchange rate between rest and stress. Our simulations showed that for a stress peak of 1 s–1, this approximation would underestimate rMBF by 6% for fast exchange. Therefore, this effect, although neglected, could in fact contribute to the underestimation of MPR in control pigs. Third, the assumption of exchange was not considered as a function of time and position along the capillary. However, since contrast agent concentration varies in time and along the length of the capillary, some portions of the intravascular space may be in fast exchange with the extravascular space, and others may be in slow or intermediate exchange at the same time (15). Simulations using the model presented in Ref. 15 could help elucidate the impact of such simplification on rMBF quantification.
tion (order of magnitude ⫽ 0.35). However, since the peak left ventricle T1 is of the same order as TI, the AIF concentration might be underestimated using signal analysis instead of ⌬R1 analysis. This effect, added to the two described above, would result in a greater overestimation of rMBF with signal analysis than with ⌬R1 analysis. Quantification of MPR using signal analysis suffered from the same limitations as ⌬R1 analysis with the fast-exchange approximation. In ischemic pigs, this technique appeared to be less accurate (i.e., it underestimated MPR) but more noise-robust (there was less dispersion, as seen in the smaller SD and a narrower confidence interval for rMBF values). Signal analysis is a more straightforward method and requires little data interaction as compared to ⌬R1 analysis, which may be an advantage in the case of low SNRs. The use of a saturation-recovery True-FISP instead of a Turbo-FLASH sequence might improve the SNR and hence the quantification of perfusion even in regions with a perfusion defect (34).
Impact of Signal Analysis vs. ⌬R1 Analysis on rMBF and MPR Quantification
CONCLUSIONS
Overestimation of rMBF by ⌬R1 analysis and the fastexchange approximation (Fig. 4) could be explained by two main factors. First, the AIF may be underestimated as a consequence of the nonlinearity of signal enhancement at high contrast agent concentrations attributable in particular to T2 effects. The overall orders of magnitude for ⌬R1 were in good agreement with blood sample measurements, but because of bolus dispersion through the lungs between the left ventricle and carotid, the blood sample curves had lower peaks and larger widths (Fig. 2) (29). Therefore, we cannot exclude the possibility of a slight underestimation of the MRI-derived AIF peak in the current study. In a recent paper Gatehouse et al. (30) proposed a sequence that allows accurate measurement of the AIF based on the use of different saturation times for the AIF (short) and myocardium (long) (31). This technique could be an alternative to the dual-bolus injection technique, which has also been proposed as a means of measuring undistorted AIF (32,33). Most importantly, simulations using an fBV of 9% and isolated heart data showed that the ARMA deconvolution technique overestimated blood flow for rMBF values lower than 250 ml/min/100 g, and underestimated blood flow above that threshold (23). However, the same study proved the ARMA method to be more robust over the myocardial flow range as a whole than singular value decomposition (SVD) or analytical deconvolution using either an exponential or a Fermi function (23). This overestimation of rMBF at high flow when using ARMA deconvolution also explains the trend seen in the bias between microsphere- and MRI-derived rMBF as flow increases. The linear relation of signal to R1 is based on TI and T1, and hence on contrast agent dose (TI/T1 ⬍⬍ 1). The contrast agent used in the present study, CMD-A2-Gd-DOTA, had a high T1-relaxivity in blood (9.4 mM–1.s–1 at 1.5 T) and the injected dose (0.009 mmol Gd/kg) was selected to reach the contrast-to-noise ratio (CNR) necessary for absolute quantification (23). At peak tissue concentration, TI/T1 was less than 1 under the fast-exchange approxima-
The present study demonstrated in a small population of control and ischemic pigs that the fast-exchange approximation can be appropriate for estimating rMBF using a contrast-enhanced saturation-recovery Turbo-FLASH sequence with a TI of 176 ms and an intravascular contrast agent. Such a protocol could provide a good compromise for performing both visual assessment of ischemicallyinjured tissue and absolute quantification of rMBF. The contrast agent dose should be carefully selected so as to maintain a linear relationship between contrast agent concentration and ⌬R1 in the blood, especially for novel T1 blood-pool agents with high relaxivity, as in the present study. Anticipating official approval of a macromolecular contrast agent with prolonged vascular retention, the clinical utility of this technique will require additional clinical testing in patients (33). When precontrast T1 quantification is not feasible (for example, because of limited scanning time in clinical practice), signal analysis using the current protocol, assuming a linear relationship between signal and concentration, may also prove valuable. ACKNOWLEDGMENTS This research was supported in part by Guerbet and Claude Bernard Lyon I University. Sabin Carme was supported by the French Ministry of Research and Technology (grant 00-2-11922) and by a grant from the Rhoˆne Alpes regional authority (Scholarship Eurodoc #0100898801). The authors thank Prof. Marc Janier and Dr. Alejandro Mazzadi from Creatis UMR CNRS 5515 for fruitful discussions and valuable contributions to the experiments. REFERENCES 1. Canet EP, Casali C, Desenfant A, An MY, Corot C, Obadia JF, Revel D, Janier MF. Kinetic characterization of CMD-A2-Gd-DOTA as an intravascular contrast agent for myocardial perfusion measurement with MRI. Magn Reson Med 2000;43:403– 409. 2. Jerosch-Herold M, Seethamraju RT, Swingen CM, Wilke NM, Stillman AE. Analysis of myocardial perfusion MRI. J Magn Reson Imaging 2004;19:758 –770.
Myocardial Water Exchange and Blood Flow 3. Schwickert HC, Roberts TP, Shames DM, van Dijke CF, Disston A, Muhler A, Mann JS, Brasch RC. Quantification of liver blood volume: comparison of ultra short TI inversion recovery echo planar imaging (ULSTIR-EPI), with dynamic 3D-gradient recalled echo imaging. Magn Reson Med 1995;34:845– 852. 4. Koenig SH, Spiller M, Brown 3rd RD, Wolf GL. Relaxation of water protons in the intra- and extracellular regions of blood containing Gd(DTPA). Magn Reson Med 1986;3:791–795. 5. Bauer WR, Roder F, Hiller KH, Han H, Frohlich S, Rommel E, Haase A, Ertl G. The effect of perfusion on T1 after slice-selective spin inversion in the isolated cardioplegic rat heart: measurement of a lower bound of intracapillary-extravascular water proton exchange rate. Magn Reson Med 1997;38:917–923. 6. Donahue KM, Burstein D, Manning WJ, Gray ML. Studies of Gd-DTPA relaxivity and proton exchange rates in tissue. Magn Reson Med 1994; 32:66 –76. 7. Donahue KM, Weisskoff RM, Burstein D. Water diffusion and exchange as they influence contrast enhancement. J Magn Reson Imaging 1997; 7:102–110. 8. Schwarzbauer C, Morrissey SP, Deichmann R, Hillenbrand C, Syha J, Adolf H, Noth U, Haase A. Quantitative magnetic resonance imaging of capillary water permeability and regional blood volume with an intravascular MR contrast agent. Magn Reson Med 1997;37:769 –777. 9. Cao Y, Brown SL, Knight RA, Fenstermacher JD, Ewing JR. Effect of intravascular-to-extravascular water exchange on the determination of blood-to-tissue transfer constant by magnetic resonance imaging. Magn Reson Med 2005;53:282–293. 10. Judd RM, Reeder SB, May-Newman K. Effects of water exchange on the measurement of myocardial perfusion using paramagnetic contrast agents. Magn Reson Med 1999;41:334 –342. 11. Larsson HB, Rosenbaum S, Fritz-Hansen T. Quantification of the effect of water exchange in dynamic contrast MRI perfusion measurements in the brain and heart. Magn Reson Med 2001;46:272–281. 12. Wendland MF, Saeed M, Yu KK, Roberts TP, Lauerma K, Derugin N, Varadarajan J, Watson AD, Higgins CB. Inversion recovery EPI of bolus transit in rat myocardium using intra- and extravascular gadoliniumbased MR contrast media: dose effects on peak signal enhancement. Magn Reson Med 1994;32:319 –329. 13. Jerosch-Herold M, Wilke N, Wang Y, Gong GR, Mansoor AM, Huang H, Gurchumelidze S, Stillman AE. Direct comparison of an intravascular and an extracellular contrast agent for quantification of myocardial perfusion. Cardiac MRI Group. Int J Card Imaging 1999;15:453– 464. 14. Lombardi M, Jones RA, Westby J, Torheim G, Southon TE, Haraldseth O, Michelassi C, Kvaerness J, Rinck PA, L’Abbate A. Use of the mean transit time of an intravascular contrast agent as an exchange-insensitive index of myocardial perfusion. J Magn Reson Imaging 1999;9:402– 408. 15. Moran GR, Prato FS. Modeling (1H) exchange: an estimate of the error introduced in MRI by assuming the fast exchange limit in bolus tracking. Magn Reson Med 2004;51:816 – 827. 16. Bjornerud A, Bjerner T, Johansson LO, Ahlstrom HK. Assessment of myocardial blood volume and water exchange: theoretical considerations and in vivo results. Magn Reson Med 2003;49:828 – 837. 17. Vallee JP, Ivancevic M, Lazeyras F, Kasuboski L, Chatelain P, Righetti A, Didier D. Use of high flip angle in T1-prepared FAST sequences for myocardial perfusion quantification. Eur Radiol 2003;13:507–514. 18. Carme S, Mai W, Mazzadi A, Janier M, Revel D, Magnin I, Wiart M, Neyran B, Canet Soulas E. Importance of parametric mapping and deconvolution in analyzing MR myocardial perfusion images. Invest Radiol 2006 ;41:374 –383.
347 19. Hazlewood CF, Chang DC, Nichols BL, Woessner DE. Nuclear magnetic resonance transverse relaxation times of water protons in skeletal muscle. Biophys J 1974;14:583– 606. 20. Larsson HB, Fritz-Hansen T, Rostrup E, Sondergaard L, Ring P, Henriksen O. Myocardial perfusion modeling using MRI. Magn Reson Med 1996;35:716 –726. 21. Bauer WR, Hiller KH, Roder F, Rommel E, Ertl G, Haase A. Magnetization exchange in capillaries by microcirculation affects diffusion-controlled spin-relaxation: a model which describes the effect of perfusion on relaxation enhancement by intravascular contrast agents. Magn Reson Med 1996;35:43–55. 22. Cerqueira MD, Weissman NJ, Dilsizian V, Jacobs AK, Kaul S, Laskey WK, Pennell DJ, Rumberger JA, Ryan T, Verani MS. Standardized myocardial segmentation and nomenclature for tomographic imaging of the heart: a statement for healthcare professionals from the Cardiac Imaging Committee of the Council on Clinical Cardiology of the American Heart Association. Circulation 2002;105:539 –542. 23. Neyran B, Janier MF, Casali C, Revel D, Canet Soulas EP. Mapping myocardial perfusion with an intravascular MR contrast agent: robustness of deconvolution methods at various blood flows. Magn Reson Med 2002;48:166 –179. 24. Bland JM, Altman DG. Measuring agreement in method comparison studies. Stat Methods Med Res 1999;8:135–160. 25. Gould KL, Lipscomb K. Effects of coronary stenoses on coronary flow reserve and resistance. Am J Cardiol 1974;34:48 –55. 26. Jerosch-Herold M, Wilke N, Stillman AE. Magnetic resonance quantification of the myocardial perfusion reserve with a Fermi function model for constrained deconvolution. Med Phys 1998;25:73– 84. 27. Wilke N, Jerosch-Herold M, Wang Y, Huang Y, Christensen BV, Stillman AE, Ugurbil K, McDonald K, Wilson RF. Myocardial perfusion reserve: assessment with multisection, quantitative, first-pass MR imaging. Radiology 1997;204:373–384. 28. Al-Saadi N, Nagel E, Gross M, Schnackenburg B, Paetsch I, Klein C, Fleck E. Improvement of myocardial perfusion reserve early after coronary intervention: assessment with cardiac magnetic resonance imaging. J Am Coll Cardiol 2000;36:1557–1564. 29. Fritz-Hansen T, Rostrup E, Larsson HB, Sondergaard L, Ring P, Henriksen O. Measurement of the arterial concentration of Gd-DTPA using MRI: a step toward quantitative perfusion imaging. Magn Reson Med 1996;36:225–231. 30. Gatehouse PD, Elkington AG, Ablitt NA, Yang GZ, Pennell DJ, Firmin DN. Accurate assessment of the arterial input function during highdose myocardial perfusion cardiovascular magnetic resonance. J Magn Reson Imaging 2004;20:39 – 45. 31. Nielsen G, Fritz-Hansen T, Dirks CG, Jensen GB, Larsson HB. Evaluation of heart perfusion in patients with acute myocardial infarction using dynamic contrast-enhanced magnetic resonance imaging. J Magn Reson Imaging 2004;20:403– 410. 32. Kostler H, Ritter C, Lipp M, Beer M, Hahn D, Sandstede J. Prebolus quantitative MR heart perfusion imaging. Magn Reson Med 2004;52: 296 –299. 33. Christian TF, Rettmann DW, Aletras AH, Liao SL, Taylor JL, Balaban RS, Arai AE. Absolute myocardial perfusion in canines measured by using dual-bolus first-pass MR imaging. Radiology 2004;232:677– 684. 34. Fenchel M, Helber U, Simonetti OP, Stauder NI, Kramer U, Nguyen CN, Finn JP, Claussen CD, Miller S. Multislice first-pass myocardial perfusion imaging: comparison of saturation recovery (SR)-TrueFISP-twodimensional (2D) and SR-TurboFLASH-2D pulse sequences. J Magn Reson Imaging 2004;19:555–563.
