|Home | About | Journals | Submit | Contact Us | Français|
The objective of this study was to investigate if dynamic contrast-enhanced magnetic resonance imaging (MRI) can be used to quantitate the distribution volume (ve) in regions of normal and infarcted myocardium. ve reflects the volume of the extra-cellular, extra-vascular space within the myocardial tissue. In regions of the heart where an infarct has occurred, the loss of viable cardiac cells results in an elevated ve compared to normal regions. A quantitative estimate of the magnitude and spatial distribution of ve is significant because it may provide information complementary to delayed enhancement MRI alone.
Using a hybrid gradient echo-echo planar imaging (GRE-EPI) pulse sequence on a 1.5T MRI scanner, 12 normal subjects and four infarct patients were imaged dynamically, during the injection of a contrast agent, to measure the regional blood and tissue enhancement in the left ventricular (LV) myocardium. Seven of the normal subjects and all of the infarct patients were also imaged at steady-state contrast enhancement to estimate the steady-state ratio of contrast agent in the tissue and blood (Ct/Cb)—a validated measure of ve. Normal and infarct regions of the LV were manually selected and the blood and tissue enhancement curves were fit to a compartment model to estimate ve. Also, the effect of the vascular blood signal on estimates of ve was evaluated using simulations and in the dynamic and steady-state studies.
Aggregate estimates of ve were 23.6% ± 6.3% in normal myocardium and 45.7% ± 3.4% in regions of infarct. These results were not significantly different from the reference standards of Ct/Cb (22.9% ± 6.8% and 42.6% ± 6.3%, p=0.073). From the dynamic contrast-enhanced studies, approximately one minute of scan time was necessary to estimate ve in the normal myocardium to within 10% of the steady-state estimate. In regions of infarct, up to three minutes of dynamic data was required to estimate ve to within 10% of the steady-state ve value.
By measuring the kinetics of blood and tissue enhancement in the myocardium during an extended dynamic contrast enhanced MRI study, ve may be estimated using compartment modeling.
Delayed enhancement (DE) MRI is a widely used and powerful clinical tool for the detection of myocardial infarct. Typically, a paramagnetic contrast agent is injected into a patient to temporarily distribute in the extra-cellular, extra-vascular space (ve) in the myocardium. In regions of the heart where an infarct has occurred, the loss of viable cardiac cells results in an increased ve where the contrast agent can accumulate. This increase in contrast agent shortens the T1 of the tissue, making regions of scar tissue visible when an inversion recovery sequence set to null the normal myocardium is used (1). DE MRI has become the gold standard for non-invasively identifying the presence and extent of myocardial scarring (2–4). DE MRI can also discriminate small regions of sub-endocardial infarct from viable tissue and has shown promise for reproducible sizing of infarcts and viability assessment in serial studies (5,6).
In effect, ve directly reflects how much of an image region is comprised of extra-cellular, extra-vascular space. The remaining fraction of the image region consists of viable cells and vasculature. A quantitative estimate of the magnitude and spatial distribution of ve may provide a complementary measure of the severity and characteristics of infarcts compared to DE imaging alone. Furthermore, the accurate estimation of regional ve maps of the heart may provide a means to track changes in the degree of myocardial scarring in follow-up studies.
The steady-state (Ct/Cb) tissue enhancement method is one validated means of quantitating ve in the myocardium (7). To estimate ve using this method, a slow infusion or a bolus injection (7–10) of contrast agent is given and allowed to reach a near-equilibrium state of contrast agent transfer into and out of the tissue. At equilibrium, the ratio of the concentration of contrast agent in the vasculature to the concentration of contrast agent in the extra-cellular extra-vascular space constitutes the partition coefficient (11). When the partition coefficient is scaled to account for the blood hematocrit, then the steady-state ratio of contrast agent in the tissue and blood, Ct/Cb, directly reflects the myocardial distribution volume, ve (10,12,13). Here, Ct is the concentration of contrast agent in the tissue and Cb is the concentration of contrast agent in the blood. The bolus and slow infusion methods require nearly 10 minutes and 20 minutes (7,9,10), respectively, for the flow of contrast agent into and out of the myocardium to reach a steady-state condition.
We hypothesize that a relatively short dynamic contrast-enhanced MRI scan can also quantitate ve in the myocardium. With this proposed method, the kinetics of myocardial blood and tissue enhancement are imaged dynamically for one to five minutes after a bolus injection of contrast agent. The enhancement of the myocardium is assumed to increase and decrease according to a physiologically derived compartmental model (14,15). ve is then estimated from the kinetics of tissue enhancement in the myocardium, according to the representative model. While ve values from dynamic contrast-enhanced scans have been reported (12,16–19), they have not been validated, and the one study that included infarct regions (19) reported a decrease in ve with infarction, contrary to expectations.
Simulation and patient studies were performed in order to evaluate the accuracy of estimating ve from dynamic contrast-enhanced perfusion data. The simulation studies were used to investigate how imaging time and the inclusion of a vascular blood signal (Vb) in the kinetic and Ct/Cb models would affect ve estimates in a typical case. Preliminary patient studies were also performed to determine the minimum imaging time necessary to fully capture the delayed kinetics of tissue and blood enhancement in regions of normal and infarcted tissue. The primary goals of the patient studies were 1) to compare ve and Ct/Cb estimates in normal and infarcted regions of myocardium and 2) to investigate how well the results agreed with predictions based on simulations with the inclusion of Vb.
For the simulation studies, dynamic tissue enhancement curves, Ctis(t), were generated by the numerical convolution of a patient-derived blood enhancement curve, Cbld(t), with a modified Kety model (12,14). The representative kinetic tissue enhancement model is given as:
From this kinetic model, Ktrans and kep are rate constants of influx and efflux of contrast agent into and out of the tissue, respectively. ve is defined as the ratio of Ktrans to kep for each tissue curve, scaled by an average blood hematocrit value (Hct=0.45): ve =Ktrans/kep*(1-Hct) (10,12). The Vb term included in Ctis(t) represents the fractional component of the blood vasculature in the myocardium. The Δt term represents the time delay between LV blood enhancement and myocardial tissue enhancement, as blood passes through the coronary arteries. For the simulations of Ctis(t), physiologic values of Ktrans, kep, Vb, and Hct (12,18,19) were used in the convolution function. For normal myocardium: Ktrans=0.6 ml/g/min, kep=1.65 min−1, Vb=0.04, and Hct=0.45. For infarcted myocardium: Ktrans=0.4 ml/g/min, kep=0.5 min−1, Vb=0.04, and Hct=0.45.
The Cbld(t) curve used in the simulations was interpolated from two dynamic signal intensity (SI) curves that were acquired using a Siemens Avanto 1.5T MRI scanner (Siemens Medical Systems, Erlangen, Germany) using a hybrid GRE-EPI pulse sequence (20) with saturation recovery magnetization preparation. The first 300s portion of Cbld(t) was taken from the mean SI of a manually selected region of the the LV blood pool of a healthy subject immediately after the injection of a 0.025 mmol/kg dose of gadolinium. The final 10s of Cbld(t) was acquired from the mean SI of the same LV blood pool region after 15 minutes. The two SI segments of Cbld(t) were interpolated with three piecewise-linear line segments to construct a continuous Cbld(t) curve that was 900s long. To ensure realistic dynamics of blood enhancement between the acquired SI segments, the slopes of the interpolated segments of Cbld(t) were chosen so that the ratio of Ctis(t) to Cbld(t) was approximately constant for the entire length of the curves. For this constraint, Ctis(t) was numerically generated from Cbld(t) using Eq. (1) with influx and efflux constants typical for a healthy region of myocardium (see above). Cbld(t) was chosen to be 900s in length because tissue enhancement is nearly constant after this time duration (9,10).
While tissue enhancement curves from regions of infarct may vary dramatically from regions of healthy myocardium, a single LV blood enhancement curve is typically assumed to serve as the input to regions of both normal and scarred myocardium in infarct patients, Thus, for the limited simulations performed here, a single Cbld(t) curve was used for simulating normal and infarct regions. The analysis of the variability of Cbld(t) curves in different patients was beyond the scope of this study. Finally, prior to generating Ctis(t) from the LV blood enhancement curve, Cbld(t) was smoothed with a temporal median filter to reduce the noise in the first-pass enhancement region of the curve.
In order to estimate the amount of imaging time necessary to accurately measure ve in dynamic perfusion scans in humans, the effect of the tissue and blood enhancement curve length (the imaging time) on the estimates of ve and Ct/Cb in simulations was evaluated. Using the simulated curves described above, progressively longer portions of the “tails” of the enhancement curves were truncated, leaving less and less kinetic data to fit to the compartmental model in Eq. (1). Using a least-squares fitting algorithm, the remaining enhancement data was fit to the kinetic model to estimate ve. The estimated ve values for the truncated normal and infarcted enhancement curves were compared to the known original ve values to measure the % error between the estimate of ve and the true ve as a function of the length of the enhancement curves.
The relative steady-state change in tissue signal enhancement in the Ct/Cb model reflects the approximate change in contrast agent concentration in the myocardium, which includes the cellular and extra-cellular space and the myocardial vasculature, Vb. Because Vb may comprise on the order of 10% of the total myocardial tissue volume (18,21–23) and because the contrast agent concentration may be higher in Vb than in the remaining tissue, ve and Ct/Cb can be overestimated if the vascular blood signal contribution is not accounted for in dynamic and steady-state Ct/Cb studies. To investigate how the estimates of ve and Ct/Cb depend on Vb, we defined the steady-state Ct/Cb model in Eq. (2) to account for the physiologic Vb signal, similar to the way that Vb is included in Eq. (1). In the Ct/Cb model given below, ΔCtis*(t) is the observed change in tissue signal enhancement from pre-contrast to post-contrast with the Vb signal included. Vb was systematically changed when generating the simulated Ctis(t) curves, and the resultant changes in Ct/Cb were calculated.
All dynamic images were acquired with a Siemens Avanto 1.5T MRI scanner (Siemens Medical Systems, Erlangen, Germany) using a hybrid GRE-EPI pulse sequence (20) with saturation recovery magnetization preparation. Twelve normal subjects and four infarct patients were imaged at rest, after being given a low-dose (0.025±0.006 mmol/kg) bolus injection of Gd-DTPA (Omniscan; Amersham Health Inc., Princeton, NJ) at 6cc/s. All the infarct patients in the study had known coronary artery disease (CAD) with stable chronic infarcts with no edema present. For these patients, imaging took place from nine months to more than 25 years post-infarct. Three of the patients had non-ST-segment elevation myocardial infarcts and one had a ST-segment elevation myocardial infarct. The contrast agent bolus was followed by a saline flush of 15mL at 6cc/s using a Medrad Spectris Solaris MR power injector (Medrad, Inc., Indianola, PA) via an antecubital vein. To speed imaging, in five of the normal subjects only half of the phase encoding lines for each image were acquired, using TSENSE acceleration (24). For each subject imaged, imaging parameters were selected in order to acquire at least three short axis (SA) slices of the LV during every heart beat. Typical imaging parameters were: delay after saturation pulse=40–70ms, TR=5.8–6.5ms, TE=0.98–1.23ms, ETL=4, flip angle=20–25, 8mm slice thickness, FOV=360×270 with an image matrix=160×90.
Dynamic stress images were acquired approximately 10 minutes after the dynamic rest scans with the same sequence and parameters described above. Prior to imaging, a continuous infusion of adenosine (140 μg/kg/min) (Adenoscan; Astellas Pharma US, Inc., Deerfield, IL) was administered to each subject via an antecubital vein to induce vasodilation. Approximately three minutes after the start of the adenosine infusion, a low-dose (0.024±0.006 mmol/kg) bolus injection of Gd-DTPA followed by a saline flush of 15mL at 6cc/s was given. During the dynamic rest and stress imaging, most patients held their breath for approximately 10–20 seconds during the first-pass of contrast agent through the LV and then breathed shallowly for the remainder of the scan. At least one minute of data was obtained in all normal subjects and at least three minutes of data was acquired for infarct patients. As time and protocol permitted, longer acquisitions were performed. Figure 1 depicts a representative timeline.
In all of the healthy subjects and infarct patients, additional contrast agent was injected after the rest and stress dynamic scans were completed, for a total contrast agent dose of 0.15–0.20 mmol/kg. After a 10–15 minute wait, DE imaging was performed using a phase sensitive inversion recovery (PSIR) imaging sequence (25). The approximate inversion recovery time to null the healthy myocardium was determined subjectively by experienced cardiologists. Typical inversion times ranged from 225–275ms.
To image Ct/Cb, a series of images were acquired in seven of the normal subjects and in all four infarct patients using the GRE-EPI sequence described above. Ct/Cb imaging consisted of a scan acquired prior to the injection of any contrast agent and a post-contrast scan acquired approximately 15 minutes after the final injection of contrast agent. The pre- and post-contrast scans consisted of 10 image frames with a total image time of approximately eight to 12 seconds. The same imaging parameters and SA slices were selected to match those in the dynamic scans. The images were used to estimate Ct/Cb to give ve as described below.
All image processing was performed using Matlab Version 7.2 (The MathWorks Inc. Natick, MA). The image frames for each slice of the LV were manually registered for in-plane rigid-body motion and endocardial and epicardial contours were manually drawn in each SA slice to segment the LV myocardium for analysis. Because the pre-contrast myocardium was assumed to have a uniform signal intensity, the LV was divided into eight equiangular sections and each section was normalized to the mean pre-contrast signal of the entire LV to correct for regional MRI coil sensitivity (26,27). Dynamic tissue enhancement curves, Ctis(t), were obtained from each of the LV octants and the dynamic blood enhancement curve, Cbld(t), was obtained from the mean SI of a manually selected region of the LV blood pool in each slice. For each subject in the study, all of the LV tissue enhancement curves were simultaneously fit to the compartment model in Eq. (1) to estimate a global time delay between LV blood enhancement and myocardial tissue enhancement. This delay was then set as a fixed parameter when estimating ve for each LV region within each SA slice of the myocardium. During dynamic imaging, precise image acquisition times were obtained from the scanner in order to account for each patient’s variable heart rate during the scan. This “time-stamp” correction process ensures a more accurate estimate of kinetic values such as ve from the dynamic enhancement curves (28).
A non-linear least-squares curve fitting algorithm was used for the analysis. For each of the normal subjects in the study, approximately one minute of tissue and blood enhancement data was included in the model to estimate ve. For the infarct patients, at least three minutes of tissue and blood enhancement data was acquired for the analysis. For three of the infarct patients, five minutes of enhancement data was acquired to verify that an approximate steady-state tissue enhancement had been reached. Also, in all but one of the infarct patients, the enhancement curves were acquired from a single continuous dynamic scan. In the remaining infarct patient, three dynamic acquisitions separated by approximately one minute each were acquired and then interpolated to construct a single continuous five minute long set of enhancement data The imaging time of the blood and tissue enhancement curves was chosen based on simulations and preliminary patient studies, which revealed that ve error estimates were within 10% of the steady-state ve value when one minute of non-infarct enhancement data and between one and three minutes of infarct enhancement data was used in the curve fits.
To investigate the effects of imaging time on the kinetic model estimates of ve in humans, the dynamic perfusion data for the normal subjects and infarct patients was analyzed to determine the amount of imaging time required for the distribution of contrast agent in the myocardium to reach a “near-equilibrium” state with slowly changing kinetics. Before analysis, the acquired dynamic images were manually registered for in-plane rigid-body motion and regions of healthy and scarred myocardium were manually selected to generate average normal and infarct tissue enhancement curves. For the analysis of myocardial scar, core regions of infarct were manually identified and selected from delayed enhancement images, while peri-infarct scar was not included in the analysis. The full length acquired enhancement curves were fit to the compartmental model in Eq. (1) to estimate a “truth” value of ve for the normal and infarcted regions of myocardium.
As with the simulation studies, progressively larger portions of the “tails” of the enhancement curves were removed and the remaining data was fit the compartmental model to estimate ve. The estimates of ve for the truncated normal and infarct enhancement curves were compared to the original “truth” values of normal and infarct ve to measure the % error between the two versus different lengths of enhancement curves. This truncation process was used to determine the minimum amount of image time necessary, in both normal and infarcted regions of tissue, to estimate ve within a desired percentage of the “truth” value.
To estimate Ct/Cb in the myocardium, the 10 image frames of both the steady-state pre-contrast and the final steady-state post-contrast dynamic images were manually registered and averaged to create a mean pre-contrast and post-contrast enhancement image. The average pre-contrast and post-contrast images were then manually registered to one another. The Ct/Cb image was computed using Eq. (2) by subtracting the mean pre-contrast image from the mean post-contrast image and scaling the resultant image by removing the Vb signal contribution and then dividing the final image by the mean relative enhancement of the LV blood signal. The mean model estimate of Vb from all of the dynamic contrast-enhanced perfusion studies, Vb=0.04, was used in the calculation of Ct/Cb. In all of the steady-state Ct/Cb patient studies, the endocardial and epicardial boundaries of the LV myocardium and the eight equiangular LV regions were selected to match the regions in the dynamic perfusion images. Also, for the infarct patients, regions of scar in the Ct/Cb images were manually selected (based on DE image comparison) in order to compute average Ct/Cb values for infarct regions of the myocardium without partial volume effects of the surrounding normal tissue.
Aggregate normal and infarct ve estimates in the human studies were computed as the mean and standard deviation of ve from each LV octant region used in the model curve fitting process. When evaluating the effect of the length of the tissue enhancement curves on ve estimates, the error was calculated as the % difference from the steady-state value of ve at t=900s (in the simulation studies) and the % difference from the “steady-state” value of ve (using all the enhancement data) in the human studies. Ct/Cb estimates in the simulation studies were computed using Eq. (2) and reflect the steady-state “truth” value of Ct/Cb at t=900s with the inclusion of Vb=0.04.
In the human studies, Ct/Cb in each LV octant region reflected the mean value of Ct/Cb for all the image pixels in the region. For aggregate results, the mean and standard deviation of normal and infarct Ct/Cb values from all the LV octant regions were reported. Comparative results of ve and Ct/Cb were reported as mean and standard deviation values for the seven normal subjects and the four infarct patients that had matched ve and Ct/Cb analysis. A correlation plot and a Bland-Altman plot were used to compare regional ve and Ct/Cb estimates in the patient studies and a Student’s t-test was performed to determine whether the estimates of ve and Ct/Cb were significantly different (p<0.05 was considered statistically significant).
In simulation studies of normal and infarct tissue enhancement, the effects of imaging time on the kinetic model estimates of ve were evaluated. Figure 2 shows simulated blood and tissue enhancement curves that were extended to a “steady-state” enhancement of 15 minutes. Approximately 30 seconds of the noise-free simulated kinetic tissue and blood enhancement data was required after contrast agent injection to estimate normal values of ve to within 5% of the true 900s steady-state value. For the infarct curves, approximately 1.5 minutes of simulated kinetic tissue and blood enhancement data were required to estimate ve to within 5% of truth. As more kinetic data was included to extend the “tails” of the enhancement curves, the model estimates of ve did not change significantly. Figure 3 shows the relationship between the error of the ve estimates versus time for simulations of normal and infarct tissue enhancement curves. These results are for this case only, and changes in the input function and kinetic parameters are expected to alter this relationship.
Next, the effects of Vb on steady-state Ct/Cb estimates were evaluated in the simulation studies. Figure 2 shows simulated tissue enhancement curves with and without Vb. In both normal and infarct tissue enhancement curves, the exclusion of modeling Vb resulted in decreased tissue enhancement and elevated estimates of Ct/Cb at steady-state enhancement. This implies that, if the vascular blood signal is not accounted for in modeling the flow of contrast agent into and out of the tissue, ve will be overestimated. Ct/Cb was calculated using the simulated values of Vb=0.04, and Ctis(t) and Cbld(t) at t=900s in Eq. (2). For the simulated normal tissue curves, the initial known value of ve was 0.20. The estimates of normal Ct/Cb with and without the inclusion of Vb in the model were 0.21 (3% greater than the known ve) and 0.25 (23% greater than the known ve), respectively. For the simulated infarct curves, the initial known value of ve was 0.45. The model estimates of infarct Ct/Cb with and without the inclusion of Vb were 0.46 (3% greater than the known ve) and 0.50 (12% greater than the known ve), respectively.
For the quantitation of ve in the human studies, Figures 4a and 4b show histograms of the aggregate ve estimates in all the normal subjects from the rest and stress studies. Normal values of ve at rest and stress were estimated to be 23.6% ± 6.3% and 23.9% ± 4.3%, respectively. The estimates of ve at rest and stress were not significantly different in normal subjects (p=0.12). Infarct values of ve at rest were estimated to be 45.7% ± 3.4% using three to five minutes of dynamic tissue enhancement data. The differentiation of ve in normal and infarcted myocardium demonstrates the potential of dynamic imaging to distinctly identify regions of infarct and quantify the extent of increase of ve within the myocardium. In this study, the estimates of normal and infarct ve in the study are significantly different (p=0.0001).
Figure 5 shows the histogram of Ct/Cb measurements in seven normal subjects. Aggregate estimates of Ct/Cb were 22.9% ± 6.8% and 42.6% ± 6.3% in regions of normal and infarcted myocardium, respectively. Figure 6 shows the correlation between ve and Ct/Cb in the normal subjects and infarct patients in the study. Figure 7 shows a Bland-Altman plot of the correlation between normal ve and Ct/Cb results for the seven subjects that were imaged dynamically and at steady-state. The mean Bland-Altman difference has a bias of 1.5%. A summary of myocardial ve estimates from the present study and other studies is given in Table 1. When λ was reported, a myocardial density of 1g/ml was assumed and the results were scaled by an average hematocrit of 0.45, where ve =λ*(1-Hct).
Figure 8 shows the decreasing error estimates of ve for a normal and infarcted region of myocardium in a single subject, as more dynamic contrast-enhanced data is included in the model fits. From the subjects in this study, approximately one minute of dynamic data is required to estimate ve to within 10% of the steady-state value in normal tissue regions. Between one and three minutes of dynamic data is required to estimate ve to within 10% of the steady-state value in infarcted tissue regions. The variations in the estimates of Vb from the kinetic modeling was 0.04 ± 0.03. For the Ct/Cb calculations using Eq. (2), a constant value of 0.04 was used, which resulted in Ct/Cb estimates of 22.9% ± 6.8%. When the range including one standard deviation (± 0.03) is included in the calculations of Ct/Cb, the mean ve estimates could theoretically extend from 19.9% to 25.9%.
Figure 9 shows a typical blood enhancement curve (AIF) with the corresponding tissue enhancement curves from a normal and infarcted region of the LV myocardium. Note the suppressed peak enhancement and the decreased rate of contrast agent efflux, kep, in the infarct tissue curve. Figure 10a depicts a DE image with a small infarct (bright region) in the anterio-septal region of the LV. Figure 10b shows one frame of a slice-matched, dynamic contrast-enhanced perfusion image corresponding to the DE image in Figure 10a. Figures 10c and 10d show the corresponding Ktrans and ve maps derived from kinetic modeling and Figure 10e shows the ve map derived from the steady-state Ct/Cb method.
The primary finding of this study is that estimates of myocardial ve can be quantified from dynamic contrast-enhanced perfusion MRI, using compartment models. One minute of dynamic tissue enhancement data is necessary to estimate ve in normal regions of the myocardium and three minutes of tissue enhancement data is necessary to estimate ve in regions of infarct. While only one minute of dynamic enhancement data was sufficient to estimate infarct ve to within 5% of the steady-state ve value in one infarct patient, the other patients required up to three minutes to be within 5% of the steady-state ve estimate. A larger population is needed to determine if there is a correlation between patient characteristics or the types of infarcts imaged and the amount of imaging time required to measure the steady-state ve value in regions of infarct.
The estimates of ve and Ct/Cb are not significantly different (p=0.073) in viable and scarred myocardium when the vascular blood signal, Vb, is included in both the kinetic and steady-state models. One limitation of the study is that when computing ve and Ct/Cb, the vascular blood volume was assumed to be a constant Vb=0.04, which was the mean model estimate of Vb from all of the dynamic contrast-enhanced perfusion studies. While this value of Vb is comparable to values found by another group using MRI (18,29), it is lower than estimates of Vb found by some histopathological estimates (30,31), and it may be possible that within a region of myocardium the fraction of tissue comprised of blood vasculature may not be spatially uniform.
In this study, estimates of ve and Ct/Cb are both approximately 23±6% in normal myocardium and approximately 45±6% in infarcted myocardium. These results parallel the findings from numerous studies that have measured ve in normal and infarcted myocardium, although there are some discrepancies. Table 1 summarizes these previous findings from other groups. Specifically, the estimates of ve from this study are similar to ve results from autoradiographic imaging and histological sectioning techniques (29–31) and are comparable to ve estimates from the steady-state Ct/Cb method, using regional changes in signal enhancement (7,9,10,32). ve measurements from histological studies range from 19%–30% depending on the method of chemical fixation and histological analysis. ve estimates from the Ct/Cb studies range from 16%–26% in normal myocardium and from 36%–48% in infarcted myocardium. Differences may vary according to whether the ve of acute or chronic infarct was measured or whether Vb was accounted for in the Ct/Cb model. In the current study, only stable chronic infarcts were imaged and Vb was included in the calculation of Ct/Cb.
Estimates of normal ve from several kinetic modeling studies are slightly lower than the results presented here, ranging from 10%–20% in the normal myocardium (12,16–19), and are generally lower than histology and Ct/Cb measurements. It is hypothesized that the difference in these results may be from the techniques used to convert the measured MR signal to contrast agent concentration, or because dynamic contrast-enhanced data was acquired for too short a time to accurately measure the delayed kinetics of tissue enhancement. Differences may also be due to the exclusion of the vascular blood signal in the compartmental models.
In two studies using inversion recovery sequences and T1-mapping to estimate Ct/Cb, ve results are higher than histological findings and the results of the current study. One author suggests that the elevation in ve estimates may be due to partial volume effects of mixed blood-myocardium voxels that resulted from the limited spatial resolution of the imaging sequence (11) (although other studies report similar spatial resolutions and different estimates of ve). Another author suggests the elevated measurements may be due to long-standing severe heart failure with LV remodeling in their patient population (9). The exclusion of Vb in these Ct/Cb models may also have elevated the estimates of ve. From Eq. (2), it is clear that the inclusion of Vb in the steady-state model can only reduce the estimates of Ct/Cb from the steady-state images. Similarly, from Eq. (1), the inclusion of Vb in the compartmental model can only reduce ve estimates.
Of the studies in Table 1 that included patients with infarct, all but one reported an increase in ve in regions of myocardial scarring. It is hypothesized that in regions of the heart where an infarct has occurred, the loss of viable cardiac cells results in an increased distribution volume. Furthermore, (29) reports that ve increases with increased time of coronary artery occlusion and that ve is greater in the core of an infarct than in the peripheral infarct regions. This finding suggests that there may be a range of ve values within an infarct that vary according to the severity of cellular damage in the region. Thus the one study that reported a decrease in ve in regions of infarct (19) is difficult to rationalize. The reduction in ve may be due to the short time duration of the dynamic contrast-enhanced perfusion scan or the no-reflow phenomenon (19).
Sources of inaccuracy of the ve estimates in this study include the low SNR of the dynamic contrast-enhanced images, and missed images during the dynamic scans due to poor ECG gating in some subjects. Gating inaccuracies were minimized by using the acquisition time of each image to correct for non-uniform or spurious time sampling errors in the model fitting algorithm (28). Another source of error in the estimates is patient motion during the scan (respiratory or otherwise), which can corrupt the images with out-of-plane motion and make the image segmentation process and model curve-fitting less accurate.
Accurately measuring ve may provide complementary information to DE imaging to objectively measure the spatial distribution or severity of scarring in infarct patients. Another practical advantage of including quantitative perfusion imaging in a DE study is that the ve images are intrinsically registered to the perfusion images for direct comparison or subsequent analysis. This may better allow for classifying perfusion as being within-infarct or peri-infarct ischemia. Alternatively, ve could be used to improve the specificity of perfusion imaging in the same way DE imaging has been used (33), but with better registration that matches cardiac phase and heart rate. This could better discriminate between ischemia and artifacts.
This study has demonstrated that dynamic contrast-enhanced imaging can be used to accurately quantitate ve in normal and scarred myocardium. The estimates of ve and Ct/Cb are not significantly different (p=0.073) in human and simulation studies, when three minutes of post-injection dynamic contrast-enhanced data is acquired in regions of infarct. One minute of dynamic contrast-enhanced data may be sufficient to estimate ve if the patient has no myocardial scarring. To ensure that ve and Ct/Cb estimates are not over-estimated, the dynamic and steady-state models must account for the signal intensity increase from contrast agent in the myocardial vasculature, Vb. Additional simulation and patient studies are in progress to evaluate the utility of ve quantitation in research and clinical settings, including pathophysiology beyond chronic scar.
This work was supported by NIH R01 EB00177
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.