The objective of this study was to determine the feasibility of delayed-enhancement micro-computed tomography (μCT) imaging to quantify myocardial infarct size in experimental mouse models. A total of 20 mice were imaged 5 or 35 days after surgical ligation of the left coronary artery or sham surgery (n = 6 or 7 per group). We utilized a prototype μCT that covers a three-dimensional (3D) volume with an isotropic spatial resolution of 100 μm. A series of image acquisitions were started after a 200 μl bolus of a high-molecular-weight blood pool CT agent to outline the ventricles. CT imaging was continuously performed over 60 min, while an intravenous constant infusion with iopamidol 370 was started at a dosage of 1 ml/h. Thirty minutes after the initiation of this infusion, signal intensity in Hounsfield units was significantly higher in the infarct than in the remote, uninjured myocardium. Cardiac morphology and motion were visualized with excellent contrast and in fine detail. In vivo CT determination of infarct size at the midventricular level was in good agreement with ex vivo staining with triphenyltetrazolium chloride [5 days post-myocardial infarction (MI): r2 = 0.86, P < 0.01; 35 days post-MI: r2 = 0.92, P < 0.01]. In addition, we detected significant left ventricular remodeling consisting of left ventricular dilation and decreased ejection fraction. 3D cine μCT reliably and rapidly quantifies infarct size and assesses murine anatomy and physiology after coronary ligation, despite the small size and fast movement of the mouse heart. This efficient imaging tool is a valuable addition to the current phenotyping armamentarium and will allow rapid testing of novel drugs and cell-based interventions in murine models.
- computed tomography
small-animal models of cardiac disease, and in particular mouse models, play a central role in cardiovascular research and have proven their value and clinical translatability. For instance, angiotensin-converting enzyme inhibitors, which were initially developed for the treatment of hypertension, were first assessed in rats with coronary ligation (24) and subsequently in mice (30). In addition, transgenic mouse models have shed light on infarct healing and left ventricular (LV) remodeling, and are often the primary and first-line animal model for stem cell research for myocardial repair (1, 7). LV dilation and hypertrophy are often assessed ex vivo; however, the value of noninvasive imaging is being increasingly appreciated, since it allows serial imaging at high sensitivity. The assessment of myocardial function in mice is challenging because of the small size of their hearts (∼5 mm in cross diameter) and the rapid rates at which they beat (∼600 beats/min). To date, the most frequently applied imaging modalities in murine models are echocardiography and magnetic resonance imaging (MRI). The advantages of cardiac MRI include its tomographic nature, high spatial and temporal resolution, and excellent soft tissue contrast. In addition, delayed-enhancement imaging after the administration of gadolinium chelates allows determination of myocardial viability and infarct size, which are both powerful predictors of LV remodeling and heart failure (12, 20, 29). Dedicated small-animal cardiac MRI, however, is expensive and still not widely available. Many studies have therefore utilized echocardiography, at the expense of sensitivity. Grothues et al. (14) estimated that the number of patients to be examined for detecting a 10% change in LV mass is 10 times higher for echocardiography compared with MRI. These data likely also apply to cardiac imaging in small animals.
The development of cardiac micro-computed tomography (μCT) imaging in mice could thus provide the investigator with a second tomographic technique with which to investigate cardiac pathophysiology. Recently, Badea et al. (5) described the use of cine three-dimensional (3D) μCT, and contrast-enhanced CT has been used to predict myocardial viability in porcine and canine models of myocardial infarction (MI) (13, 17). However, there is a general lack of information on how well μCT can visualize infarcted myocardium in mice. We therefore hypothesized that μCT can quantify infarct size by delayed enhancement in living mice similarly as in large-animal models and in patients (13). Furthermore, we tested the ability of cine 3D μCT to detect LV remodeling at later time points after MI.
MATERIALS AND METHODS
Mouse model of MI.
A total of 22 female C57BL/6 mice of a mean age of 15 wk were used in this study. MI was induced by coronary ligation at the Center for Molecular Imaging Research surgery core, and the mortality of this procedure was 0%. Mice were anesthetized for surgery by inhalation anesthesia (isoflurane 1–2% vol/vol + 2 liters O2). Mice were intubated and ventilated with isoflurane supplemented with oxygen. The chest wall was shaved, and a thoracotomy was performed in the fourth left intercostal space. After insertion of a retractor, the LV was visualized and the left coronary artery was permanently ligated with a monofilament nylon 8-0 suture at the site of its emergence from under the left atrium. This procedure resulted in variable MIs involving the anterolateral, posterior, and apical regions and typically spared the septum (16). For smaller infarcts, we chose a more apical ligation site. Finally, the thoracotomy was closed with sutures and superglue. After recovery, mice were transferred to the Center for In Vivo Microscopy, Duke Medical Center, for CT imaging. Two mice with large infarcts died during induction of anesthesia before CT imaging. Therefore, CT imaging was performed in seven noninfarcted mice, six mice were imaged on day 5 after MI, and seven mice were imaged on day 35 after MI. The institutional subcommittee on research animal care at the Massachusetts General Hospital and the Duke University Institutional Animal Care and Use Committee approved all animal surgery and imaging studies.
Animal preparation for CT imaging.
Mice were anesthetized with ketamine (115 mg/kg) and diazepam (27 mg/kg) and intubated for mechanical ventilation at a rate of 90 breaths/min with a tidal volume of 0.4 ml. A pressure transducer on the breathing valve measured airway pressure, and electrodes taped to the animal footpads acquired the ECG signal (Blue Sensor, Medicotest). Both signals were processed with Coulbourn modules (Coulbourn Instruments, Allentown, PA) and displayed on a computer monitor with a custom-written LabVIEW application (National Instruments, Austin, TX). Body temperature was recorded and maintained at 36.5°C by an infrared lamp and feedback controller system (Digi-Sense, Cole Parmer, Chicago, IL). A catheter was inserted in the tail vein and used for the delivery of contrast agents. Animals were placed in a cradle and scanned in a vertical position. During imaging we maintained anesthesia with 1–2% isoflurane (Halocarbon, River Edge, NJ).
We chose a dual-contrast agent strategy that involved a macromolecular blood pool agent to initially visualize cardiac morphology and, in a second step, a small-molecule contrast agent that extravasates and enhances the infarct. The imaging protocol in Fig. 1 summarizes the timing of the contrast agent application: 1) the blood pool contrast agent Fenestra VC (ART Advanced Research Technologies, Saint-Laurent, QC, Canada) and 2) the conventional contrast agent iopamidol 370 (Isovue, Bracco Diagnostics, Princeton, NJ). A single bolus of Fenestra (0.01 ml/g) was applied before imaging to ensure an initial contrast between the blood and the myocardium, which allowed us to place exact regions of interest (ROIs) in the LV wall and the cavity. After a first set of image acquisitions, the conventional contrast agent iopamidol was delivered intravenously at a rate of 1 ml/h with an infusion pump (Pegasus, Instech Solomon, Plymouth Meeting, PA) with a constant infusion to hyperenhance the infarcted myocardium. In preliminary experiments, we tested this infusion rate and found that this volume is well tolerated even by mice compromised with heart failure due to extensive MI.
CT image acquisition.
We used a prototype cone beam μCT system with a design that overcomes two significant barriers to μCT in small animals (5): the reduced signal-to-noise ratio (SNR) imposed by the smaller voxels and the rapid murine heart motion at a rate of ∼600 beats/min. The system consists of a high-flux rotating anode X-ray tube designed for clinical angiography (Philips SRO 09 50, Philips Medical Systems, Bothell, WA) with a dual focal spot operating at 9 kW (0.3-mm focal spot) or 50 kW (1.0-mm focal spot). The detector is a cooled charge-coupled device camera with a Gd2O2S phosphor on a 3:1 fiber-optic reducer (X-ray ImageStar, Photonic Science, Robertsbridge, UK). The camera has a 106-mm2 active field of view with an image matrix of 2,048 × 2,048 and a resulting pixel size of 51 × 51 μm. The tube and the detector are mounted horizontally on aluminum frame components (80/20, Bellevue, WA), and the animal is held in a vertical position centered in front of the detector face. The distance between the detector and the animal is 40 mm and between the animal and the X-ray source 480 mm. This configuration results in geometric blur of the focal spot that matches the Nyquist sample at the detector (3). Thus this geometry provides magnification of 1.09, essentially eliminating penumbral blur. The use of a fixed gantry with a rotating animal stage has two distinct benefits. It permits the use of larger focal spots, which provides for a higher radiation flux. Also, this geometry also allows a step-and-shoot rotation with arbitrary intervals, which facilitates synchronization of the exposure to the heart cycle. The system takes advantage of integrated physiological monitoring including respiratory and cardiac motion to limit artifacts and blur. Acquisition parameters were 80 kVp, 200 mA, and 9-ms exposures per projection. Typically, a full-image data set covered 189° (i.e., 180° plus fan angle) in 0.75° steps for a total of 252 projections. We covered the heart cycle with eight scans using prospective cardiorespiratory gating, yielding a temporal resolution of ∼15 ms. Total acquisition time for a single dataset was 7–8 min. This high-resolution 3D data set from different time points of the cardiac cycle was then displayed as a cine movie (3D cine μCT), allowing us to quantify heart function. Acquisition of the cine data set was started immediately after the beginning of the iopamidol infusion. Dosimetric measurements were performed with a Wireless Dosimetry System Mobile MOSFET TN-RD-16, SN 63 B (Thomson/Nielsen, Ottawa, ON, Canada). Five MOSFET dosimeter silicon chips (active area 0.2 mm × 0.2 mm) were positioned at the surface and the center of a acrylic rodentlike phantom. During image acquisition, the measured dose for the 3D data set was 15.4 cGy for a single time point and 1.23 Gy for eight time points. Projections were used to reconstruct tomograms with a Feldkamp algorithm using Parker weighting (10) with a Cobra EXXIM software package (EXXIM Computing, Livermore, CA). Data sets were reconstructed as isotropic 5123 arrays with effective digital sampling in the image plane of 100 μm.
The enhancement over time was measured with ImageJ (National Institute of Mental Health). Regions of interest of ∼40 pixels were selected in the LV cavity, remote, viable myocardium, infarcted lateral wall, and skeletal muscle to calculate mean signal intensities in Hounsfield units. Comparison of the infarct size to triphenyltetrazolium chloride (TTC) staining was performed on the CT images acquired 60 min after the start of iopamidol infusion, and all data sets were utilized for volumetric quantification employing a novel volumetric analysis method integrated into the Pittsburgh Supercomputing Center Volume Browser (PSC-VB) (http://www.psc.edu/biomed/research/VB). The semiautomated analysis recovers the mean signal intensity of both the blood and the myocardium. To accomplish this, the composite histogram is processed by a Gaussian mixture modeling approach (2). In addition to the direct measurement of LV cavity volume the process also produces a signal intensity that best reproduces the measured LV volume through a threshold segmentation of the ROI voxels. Diastole and systole were selected as the heart phases that provided the maximum and minimum LV volume. Stroke volume (SV) and ejection fraction (EF) are calculated with the end-diastolic (EDV) and end-systolic (ESV) volumes (SV = EDV − ESV; EF = SV/EDV).
All mice were euthanized immediately on completion of the imaging studies. Hearts were excised and rinsed in PBS and cut into myocardial rings of 1-mm thickness. Thereafter, midventricular sections were stained with TTC for 20 min, and digital pictures were acquired for quantification of infarct size by digital planimetry. Infarct size for each mouse was expressed as percent area unstained by TTC in relation to the LV area of the same midventricular slice. The corresponding CT images were identified within the 3D data set through anatomic landmarks such as papillary muscles with OsiriX shareware.
Results are expressed as means ± SD. Statistical comparisons among two groups were evaluated by Student's t-test. P < 0.05 was considered to indicate statistical significance.
Constant infusion of iopamidol leads to infarct hyperenhancement.
Cine 3D μCT produced high-resolution time-resolved images that facilitated detailed assessment of murine cardiac anatomy and function. Application of a 200-μl bolus of the intravascular contrast agent established blood-myocardial contrast to unambiguously distinguish cardiac anatomy; however, no hyperenhancement of the infarct was detected (Fig. 2). Conversely, 20 and 30 min after the start of iopamidol infusion the signal intensity in the infarct area was significantly higher than in the remote myocardium (Fig. 2), and the difference steadily increased up to 1 h. This finding applied to both the subacute state on day 5 after MI and the chronic infarct after 35 days. Mice were euthanized without regaining consciousness at the end of the imaging study, so no information on recovery is available.
μCT accurately assesses infarct size.
A good correlation was found between infarct size measured by CT, which was defined by hyperenhancement, and the ex vivo gold standard, TTC-stained sections (Fig. 3). The unstained, pale infarct area in the viability stain correlated closely to the hyperenhanced area on CT, both on day 5 (r2 = 0.86, P < 0.01) and on day 35 (r2 = 0.92, P < 0.01) after MI. On day 5, infarct size measured by CT was 33.9 ± 8.9% and by TTC 34.8 ± 8.3% [P = not significant (NS)] and on day 35 30.1 ± 11.6% and 30.2 ± 13.2% (P = NS), respectively.
Automatic segmentation detects hyperenhanced infarct.
We analyzed LV volumes with a semiautomated approach that relies on detection of differences in pixel signal intensities. An example of a segmented LV is shown in Fig. 4 and in the supplemental data (the online version of this article contains supplemental data). After placement of an initial volume of interest around the LV, the software supplies a histogram of signal intensities in this region (Fig. 4, B and D). Within this histogram, the threshold between compartments can now be defined interactively, while the segmentation result of the 3D slab is visualized (Fig. 4, E and F). In healthy hearts, the histogram shows two peaks of voxel numbers: one at the blood signal intensity and one at myocardial signal intensity (Fig. 4B). In hearts with MI, we observed a third peak, and this middle peak with intensity values between the blood and the myocardium was caused by the enhanced infarct (Fig. 4, C, and F). The optimum threshold automatically found by a Gaussian mixture modeling approach classified the infarct voxels as blood (volume A, see arrows in Fig. 4, C and F). This confirms the hyperenhancement and allows 3D visualization of the infarct zone. For quantification of the LV cavity volume the threshold can be manually corrected to include only the blood pool (volume B). This feature also allows us to rapidly quantify the 3D volume of the hyperenhanced MI by calculation of the difference between volumes A and B.
Cine 3D μCT detects LV remodeling after MI.
We compared LV morphology and function in sham-operated mice without MI with those in mice 5 and 35 days after coronary ligation. On day 5, some acute dilation of the LV was detected; however, this had not yet reached significance (Fig. 5E). As early as 5 days after infarction, the EF was severely diminished (Fig. 5F). Thirty-five days after coronary ligation, the LV had dilated significantly. In the mouse with the largest infarct, EDV increased to 150 μl, three times the normal value. The EF remained severely impaired (28 ± 7%, Fig. 5F).
The mouse model of myocardial infarction has been exceedingly valuable in heart failure research (18, 19, 27) and has been used in numerous studies investigating stem cells for cardiac repair (1, 6, 7). Here we describe a novel approach to phenotyping murine cardiac anatomy and physiology by cine 3D μCT and the application of late-hyperenhancement CT infarct imaging. The technique accurately reports 3D infarct size and provides all desired anatomic and functional cardiac data of the mouse heart (cine movie is displayed in online data supplement). Therefore, the technique is an attractive addition to the current mouse phenotyping armamentarium.
The small size and rapid movement of the mouse heart pose significant difficulties for imaging. Since the thickness of the LV wall is <1 mm, very high image resolution is mandatory and results in microscopic voxel sizes. At the same time, the rapid heart movement requires fast image acquisition. Acquisition of CT images of the mouse heart with an accuracy in scale to human cardiac CT (9, 22, 25) necessitates a 10-fold increase in temporal resolution and a 3,000-fold increase in spatial resolution. As the size of the voxel decreases, the dose must increase simultaneously to maintain sufficient SNRs (11). We employed a prototype cone beam μCT system with high X-ray photon fluence tubes that are 250 times brighter than the tubes on commercial μCT systems (3, 4, 23). This approach directly increases the SNR by an increase of the total fluence rate and resulted in high image quality despite microscopic voxel dimensions.
In line with recent reports describing late-enhancement CT in large-animal models (17) and patients (13), we found good agreement of hyperenhanced infarct scars and the ex vivo TTC stain (Fig. 3). The hyperenhanced areas accurately reported morphology and size of the infarcts. Thirty minutes after the start of the infusion of the contrast agent, sufficient differences in signal intensities were detected between the infarct and uninjured myocardium. After 60 min, the contrast-to-noise ratio between infarct and uninjured myocardium reached a mean of 7.2 ± 4.2. Future studies will show whether nontransmural infarctions induced by ischemia-reperfusion injury can be quantified in a similar fashion.
As seen in patients with acute heart failure, mice are very susceptible to volume overload after large MI. If a large-volume contrast agent bolus is injected, mice may die because of acute cardiac decompensation. Therefore, we developed the described less invasive strategy, which involves a constant, slow infusion of iopamidol rather than one large-volume bolus injection. Before imaging, we injected a small bolus of a blood pool CT contrast agent to facilitate exact placement of ROIs for assessment of signal intensities in the blood and the myocardium. The protocol can be simplified for future phenotyping experiments. Thirty minutes after the start of the infusion of iopamidol, a single cine 3D μCT data set may be acquired, which would comprehensively and noninvasively depict the murine cardiac phenotype including infarct size and EF. This approach will reduce the infused contrast agent volume to a value of 0.5 ml in total, which will attenuate changes in cardiac physiology seen with volume overload (21). In addition, a reduced number of CT scans will restrict the radiation exposure of the mouse and therefore facilitate serial imaging protocols. In our protocol, the radiation exposure was measured to be 1.23 Gy, which is ∼20% of the mouse LD50/30 (11). Therefore, an even higher sensitivity of the system would be desirable to reduce the radiation dose and contrast medium volume, which may have had effects on functional parameters such as EF in our study. Furthermore, the utility of this approach for performing serial studies in mice has yet to be assessed.
The mechanism of hyperenhancement is similar to delayed infarct hyperenhancement observed in MRI after injection of Gd-DPTA, which is based on an increased extracellular space and a therefore less restricted distribution of the low-molecular-weight contrast agent (15). In acute MI, the sarcomere membranes rupture because of ischemia, which enlarges the extracellular space. In the chronic state, the extracellular space is also relatively enlarged because of rarefication of cells. In the mature infarct, extracellular matrix such as collagen contributes extensively to the scar, and even the initially abundant fibroblasts rarefy (8).
Late-enhancement MRI of infarcts has been successfully performed in mice (31). The proposed CT technique may not reach the soft image contrast and the versatility of MRI (26); however, it offers several potential advantages. First, μCT scanners are less expensive than high-field MRI systems used for murine studies. Furthermore, the spatial resolution in our CT imaging study is at least as good as the common resolution used in murine cardiac MRI (18, 19, 27, 31), and even better than MRI in the z-direction. Murine cardiac MRI often utilizes two-dimensional gradient echo sequences, and the slice thickness is usually 1 mm (18, 19, 27, 31). The μCT setup allowed us to acquire a time-resolved 3D slab with a true isotropic resolution of 100 μm, which is 10 times higher than typical MRI techniques in the z-direction. However, this advantage in spatial resolution needs to be weighed against the lower temporal resolution possible with cine CT. Another advantage may be the relative ease of integration of the CT modality into hybrid imaging systems such as single-photon-emission CT (SPECT)-CT and PET-CT. These hybrid systems are utilized to overcome the inherent low resolution (1–2 mm) of the very sensitive nuclear imaging modalities and provide the source of a given molecular imaging agent in greater anatomic detail. While these systems are commercially available, they still lack the sophistication of the prototype μCT used in this study. Nevertheless, hybrid nuclear imaging systems are increasingly used to assess myocardial healing (28). Late-enhancement imaging can enhance the utility of the CT modality further, since it distinguishes between the infarct and the uninjured remote myocardium. This information is of importance, since molecular targets and biological processes differ greatly between the infarct and the remote myocardium. For instance, if stem cell migration is to be assessed with reporter gene imaging techniques (6), it is of interest if the cells home in on the infarct, which could be accurately outlined by late-enhancement CT. Furthermore, we believe that cine 3D μCT may evolve into a cost-effective and very sensitive diagnostic tool for the in vivo assessment of novel candidate drugs and for mouse phenotyping. In addition to rapidly delivering precise data, the noninvasive character should provide for serial studies of murine cardiac morphology, thereby greatly enhancing the sensitivity to detect remodeling of the LV.
In conclusion, we propose a novel imaging technique applicable to the mouse model of MI. Cine 3D late-enhancement μCT enables accurate measurement of murine cardiac morphology, exactly identifies the infarct region, and may be a valuable tool for murine cardiac phenotyping.
This work was supported by National Institutes of Health (NIH) Grants RO1-HL-078641 (R. Weissleder), UO1-HL-080731 (R. Weissleder), and R24-CA-92782 (R. Weissleder) and Donald W. Reynolds Foundation (R. Weissleder, M. Nahrendorf). Imaging was performed at the Duke Center for In Vivo Microscopy, an NIH/National Center for Research Resources (NCRR) National Biomedical Technology Resource (NCRR P41-RR-005959), with additional support from NIH Grants R24-CA-092656 and 5R01-HL-055348.
We acknowledge Peter Waterman for assistance with animal transfers and Dr. Art Wetzell from Pittsburgh Supercomputing Center, Carnegie Mellon University, for help with the PSC-VB software.
↵* M. Nahrendorf and C. Badea contributed equally to this work.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2007 by the American Physiological Society