Author + information
- Received October 17, 2008
- Revision received March 3, 2009
- Accepted March 4, 2009
- Published online August 1, 2009.
- Jonathan W. Weinsaft, MD⁎,†,⁎ (, )
- Raymond J. Kim, MD‡,
- Michael Ross, MD⁎,
- Daniel Krauser, MD⁎,
- Shant Manoushagian, BA⁎,
- Troy M. LaBounty, MD⁎,
- Matthew D. Cham, MD⁎,†,
- James K. Min, MD⁎,†,
- Kirsten Healy, MD⁎,
- Yi Wang, PhD†,
- Michele Parker, MS, RN‡,
- Mary J. Roman, MD⁎ and
- Richard B. Devereux, MD⁎
Reprint requests and correspondence:
Dr. Jonathan W. Weinsaft, Weill Cornell Medical College, 525 East 68th Street, New York, New York 10021
Objectives This study sought to compare contrast-enhanced anatomic imaging and contrast-enhanced tissue characterization (delayed-enhancement cardiac magnetic resonance [DE-CMR]) for left ventricular (LV) thrombus detection.
Background Contrast echocardiography (echo) detects LV thrombus based on anatomic appearance, whereas DE-CMR imaging detects thrombus based on tissue characteristics. Although DE-CMR has been validated as an accurate technique for thrombus, its utility compared with contrast echo is unknown.
Methods Multimodality imaging was performed in 121 patients at high risk for thrombus due to myocardial infarction or heart failure. Imaging included 3 anatomic imaging techniques for thrombus detection (contrast echo, noncontrast echo, cine-CMR) and a reference of DE-CMR tissue characterization. LV structural parameters were quantified to identify markers for thrombus and predictors of additive utility of contrast-enhanced thrombus imaging.
Results Twenty-four patients had thrombus by DE-CMR. Patients with thrombus had larger infarcts (by DE-CMR), more aneurysms, and lower LV ejection fraction (by CMR and echo) than those without thrombus. Contrast echo nearly doubled sensitivity (61% vs. 33%, p < 0.05) and yielded improved accuracy (92% vs. 82%, p < 0.01) versus noncontrast echo. Patients who derived incremental diagnostic utility from DE-CMR had lower LV ejection fraction versus those in whom noncontrast echo alone accurately assessed thrombus (35 ± 9% vs. 42 ± 14%, p < 0.01), with a similar trend for patients who derived incremental benefit from contrast echo (p = 0.08). Contrast echo and cine-CMR closely agreed on the diagnosis of thrombus (κ = 0.79, p < 0.001). Thrombus prevalence was lower by contrast echo than DE-CMR (p < 0.05). Thrombus detected by DE-CMR but not by contrast echo was more likely to be mural in shape or, when apical, small in volume (p < 0.05).
Conclusions Echo contrast in high-risk patients markedly improves detection of LV thrombus, but does not detect a substantial number of thrombi identified by DE-CMR tissue characterization. Thrombi detected by DE-CMR but not by contrast echo are typically mural in shape or small in volume.
Accurate detection of left ventricular (LV) thrombus is important as thrombus provides a substrate for embolic events and a rationale for anticoagulation. Both echocardiography (echo) and cardiac magnetic resonance (CMR) imaging can use contrast agents to improve thrombus detection. Echo contrast typically improves thrombus detection through cavity opacification, identifying thrombus based on anatomic appearance. CMR uses contrast to identify thrombus based on tissue characteristics related to avascularity. Using the technique of delayed-enhancement (DE)-CMR, which is widely used to discern viable from infarcted myocardium, thrombus can be identified by absence of contrast uptake. Because DE-CMR identifies thrombus based on tissue characteristics rather than anatomic appearance, it enables LV thrombus to be delineated from myocardium and chamber cavity irrespective of location or morphology.
DE-CMR has been well validated as an accurate technique for LV thrombus based on comparisons with pathology findings and clinical embolic events (1,2). In prior studies, DE-CMR has yielded a 2- to 3-fold improvement in thrombus detection versus noncontrast echo (2,3). However, comparative studies have been performed without routine use of echo contrast, which can improve thrombus detection (4). Additionally, indications for contrast use differ between CMR and echo. For CMR, almost all studies include DE-CMR for the primary purpose of distinguishing between viable and infarcted myocardium, meaning that virtually all patients receive contrast. For echo, consensus guidelines recommend that contrast be reserved for noncontrast studies with suboptimal image quality (5). The utility of a strategy of echo contrast administration based on a priori clinical risk for thrombus rather than noncontrast echo image quality is unknown. Because echo is widely used to screen patients with coronary disease or heart failure at risk for LV thrombus, optimization of diagnostic strategies for thrombus detection is of substantial importance.
The aims of this study were: 1) to compare multiple imaging techniques that identify thrombus based on anatomic appearance (noncontrast echo, contrast echo, and cine-CMR) to a reference of DE-CMR tissue characterization; 2) to assess whether thrombus detection by anatomic imaging varies in relation to LV geometry or thrombus morphology; and 3) to identify imaging markers for LV thrombus in an at-risk population.
The population consisted of patients at high risk for LV thrombus because of recent myocardial infarction or chronic heart failure who underwent a multimodality imaging protocol at Weill Cornell Medical College between August 2005 and November 2007. Patients were enrolled if CMR and echo were performed within a 7-day interval. All patients had contrast echo performed for the primary indication of LV thrombus assessment; 55% were clinically referred for CMR and 45% were recruited via an ongoing study of post-myocardial infarction thrombus and remodeling. For all patients, imaging was performed in accordance with a pre-defined protocol and interpreted by pre-assigned readers blinded to clinical history and results of other imaging modalities. Patients were enrolled de novo and none had participated in prior investigations concerning thrombus (1).
The imaging protocol consisted of noncontrast echo, contrast echo, cine-CMR, and DE-CMR. Contrast echo was performed irrespective of quality or findings of noncontrast echo; 60.3% of patients had echo and CMR performed on the same day, 32.2% underwent CMR at least one day before echo, and 7.4% underwent echo before CMR. Clinical data were collected, including cardiac risk factors, presence of coronary artery disease, and medication regimen. Pathology data (in patients undergoing LV reconstruction surgery within 1 month) were reviewed to verify thrombus as identified at the time of imaging.
All procedures were conducted in accordance with the institutional review board at Weill Cornell, which approved the study protocol. All prospectively recruited patients provided informed consent and the institutional review board approved use of pre-existing data for inclusion in the imaging registry.
The full protocol is available online.
CMR was performed using 1.5-T scanners (General Electric, Waukesha, Wisconsin). Cine-CMR was performed using a steady-state free precession sequence. Gadolinium was then intravenously administered (0.2 mmol/kg) to patients without contraindications (glomerular filtration rate <30 ml/min/1.73 m2) (6). DE-CMR was performed 10 min thereafter using a segmented inversion recovery sequence (7). Cine-CMR and DE-CMR images were acquired in matching planes. Short-axis imaging was contiguous throughout the LV. Long-axis images were acquired in 2-, 3- and 4-chamber orientations.
Echocardiograms were performed using commercial equipment (General Electric Vivid-7 or Siemens [Malvern, Pennsylvania] Sequoia) by sonographers who had undergone dedicated training concerning the imaging protocol. Noncontrast images were acquired in standard parasternal and apical imaging planes (5). A sonographic contrast agent (Definity, Lantheus Medical Imaging, North Billerica, Massachusetts) was then administered in accordance with manufacturer guidelines (8). Noncontrast- and contrast-enhanced images were acquired in at least 3 (2-, 3-, and 4-chamber) apical orientations.
Thrombus was identified on DE-CMR as a mass with tissue characteristics consistent with avascular tissue (1,2). DE-CMR was performed in accordance with the thrombus protocol that has been validated in prior research by our group (1). Standard DE-CMR was performed using an inversion time (TI) tailored to null viable myocardium (297 ± 31 ms, range 200 to 350 ms). On standard DE-CMR, thrombus typically had a gray etched appearance compared with black (viable) or white (infarcted) myocardium (Fig. 1A, standard TI). To further delineate thrombus, additional DE-CMR imaging was performed using a pulse sequence tailored to null avascular tissue (1). By increasing the TI to 600 ms, regions with contrast uptake appear gray and thrombus appears black (Fig. 1A, long TI). Scanner operators performed standard DE-CMR in all patients and long TI imaging in patients (72%) that tolerated additional breath-holds.
In accordance with established criteria (1,3), thrombus was distinguished from myocardial infarction with microvascular obstruction (MVO) based on: 1) surrounding structures (MVO encompassed by hyperenhanced myocardium); 2) appearance (MVO within myocardium, thrombus adjacent to LV cavity and typically involves abrupt endocardial transitions); and 3) stability of size on consecutive DE-CMR images (MVO shrinks from peripheral contrast fill-in, thrombus size stable).
For cine-CMR and echo, thrombus was diagnosed based on anatomic appearance and defined as an LV mass that was distinguishable from papillary muscles, trabeculae, chordal structures, technical artifact, or tangential views of the LV wall (1,9). Thrombus was echodense on echo, and signal intensity was similar to myocardium on cine-CMR.
Each modality was interpreted during a separate reading session by an experienced physician (CMR or echo American Heart Association/American College of Cardiology Level III) who was blinded to the results of other modalities and clinical history. Interobserver reproducibility was assessed by having a second reader independently reinterpret each modality for 30 randomly selected patients. All echoes were reviewed for image quality, which was graded using a scale combining scores for endocardial border definition (1 = poor, 2 = fair, 3 = excellent) and cavity artifacts (1 = present/obscuring full assessment, 2 = present/not preventing interpretation, 3 = absent).
Thrombus volume, morphology, and location were scored on DE-CMR; volume was quantified by planimetry. Morphology was classified as mural (borders concave, similar to surrounding endocardial contours) or intracavitary (borders protruding into the cavity, distinct from surrounding endocardial contours) (1,10). Location was classified as apical if localized to the LV apex or apical segments of other LV walls.
CMR and echo indexes were measured to determine relationships with thrombus. Noncontrast echo quantified left ventricular ejection fraction (LVEF) and chamber size based on linear dimensions (5). Cine-CMR quantified LVEF and chamber volumes based on planimetry. Regional contractility and infarction were scored using a 17-segment model. Segmental function was graded on cine-CMR (0 = normal contraction, 1 = mild hypokinesia, 2 = moderate hypokinesia, 3 = severe hypokinesia, 4 = akinesia, 5 = dyskinesia). Cine-CMR and echo were scored for presence of LV aneurysms. DE-CMR was scored for infarct size based on transmural extent of hyperenhancement (0 = none, 1 = 1% to 25%, 2 = 26% to 50%, 3 = 51% to 75%, 4 = 76% to 100%). Global infarct size was calculated by summing segmental scores weighted by the midpoint of the range of hyperenhancement and dividing by the total number of segments (11).
Continuous variables were compared using a student's t test with the Levene test for equality of variance to confirm homogeneity. Variables were also tested for skew and kurtosis; comparisons between non-normally distributed data (thrombus volumes) were based on logarithmic transformation. Limits of agreement (1.96 × the standard deviation of between-method differences) were calculated. Categorical variables were compared using chi-square or Fisher exact test. Diagnostic test performance and thrombus prevalence were compared using the McNemar test with exact binomial probability calculations. Magnitude of agreement between tests was measured using the kappa statistic (κ). Univariate and multivariable logistic regression analyses were performed to evaluate associations between imaging parameters and thrombus. Two-sided p < 0.05 was considered indicative of statistical significance. Analyses were performed using SPSS version 12.0 (SPSS Inc., Chicago, Illinois).
The population consisted of 121 patients who underwent CMR and echo within a mean interval of 0.8 ± 1.4 days. Both modalities showed advanced LV dysfunction (LVEF difference 0.5 ± 6.9% p = 0.42, limits of agreement −13.2% to 14.2%).
DE-CMR identified LV thrombus in 24 of 121 patients (20%). Pathology verification of thrombus was available in 3 patients, all of whom had thrombus by DE-CMR. As shown in Table 1, patients with thrombus did not differ significantly from those without thrombus in age, sex, or coronary artery disease risk factors. However, patients with thrombus had more severe LV dysfunction as measured by New York Heart Association functional class (p = 0.007) and the imaging parameters of LVEF (p ≤ 0.001) or aneurysmal dilation by echo (p = 0.09) or cine-CMR (p = 0.01). Patients with thrombus also had larger total and transmural infarct size as measured by DE-CMR (p ≤ 0.001).
Multivariable analyses were performed to identify imaging parameters associated with LV thrombus. Separate models were used to examine whether transmural myocardial infarction by DE-CMR was an independent marker for thrombus after adjustment for measures of chamber function and morphology derived from either cine-CMR or echo. When considering only CMR parameters (Table 2), transmural infarct size was an independent marker for thrombus after adjustment for LVEF and aneurysmal dilation. A similar relationship was seen when substituting echo-derived LVEF in the multivariate model, although echo-identified LV aneurysm was not an independent marker for thrombus (Table 2). Both models showed that the relationship between thrombus and infarction was continuous, such that likelihood of thrombus increased in proportion to infarct size with an approximate 40% increase in relative risk for every 10% increment in transmural infarction.
Diagnostic performance of anatomic imaging
Table 3 reports diagnostic performance of anatomic imaging techniques compared with a reference of DE-CMR. Echo contrast markedly improved diagnostic performance for thrombus detection in this high-risk population. Sensitivity improved nearly 2-fold versus noncontrast echo (p < 0.05), resulting in substantially higher accuracy (p < 0.01). Similar to contrast echo, cine-CMR provided markedly higher sensitivity and accuracy versus noncontrast echo (p ≤ 0.001). There were no significant differences in diagnostic indexes between contrast echo and cine-CMR, with strong agreement between modalities for the diagnosis of thrombus (κ = 0.79, p < 0.001). Diagnostic performance of echo was similar between patients grouped according to differences in testing sequence between CMR and echo as evidenced by nonsignificant differences in accuracy of contrast echo (p = 0.37) and noncontrast echo (p = 0.11) between groups.
Among the patent subgroup that underwent reinterpretation of images, interobserver agreement (reported as proportion of concordant reads and κ value) was good for noncontrast echo (25 of 30; κ = 0.52, p = 0.003) and excellent for contrast echo (29 of 30; κ = 0.87), cine-CMR (29 of 30; κ = 0.91), and DE-CMR (30 of 30; κ = 1.00) (all p < 0.001).
Predictors of improved contrast-echo thrombus detection
Qualitative and Structural Parameters
Multiple parameters were examined for markers that could identify patients who derived incremental diagnostic benefit from echo contrast. Table 4 reports image quality parameters for contrast and noncontrast echo in the overall population and in patients with thrombus detected or missed by each echo technique. In the total population, echo contrast improved endocardial border definition, cavity delineation, and overall image quality versus noncontrast echo (p < 0.001). However, echo performance was not associated with reader-assigned quality, with similar values for all qualitative parameters between contrast and noncontrast echoes that missed thrombus versus those that detected DE-CMR evidenced thrombus (p = NS). Similar relationships were shown when comparing image quality scores between noncontrast echoes that independently detected thrombus and those in which contrast use provided incremental utility for thrombus detection (p = NS).
Regarding LV quantitative parameters, noncontrast echoes that correctly identified or excluded thrombus had less depressed systolic function compared with echoes that were discordant with DE-CMR. As shown in Table 5, both cine-CMR and echo showed that LVEF was lower (p < 0.05) among patients with noncontrast echoes that were discordant with the DE-CMR diagnosis of thrombus. Similarly, echo-evidenced LVEF tended to be lower in patients in whom contrast echo improved thrombus assessment versus those in whom noncontrast echo alone accurately assessed thrombus (34.2 ± 9.9% vs. 41.2 ± 13.3%, p = 0.08). There were no significant differences in clinical parameters between patients with thrombus detected and those with thrombus missed by noncontrast echo.
Despite the marked overall improvement in thrombus detection yielded by a routine strategy of contrast use, 39% of DE-CMR evidenced thrombi were not detected by contrast echo, resulting in lower overall prevalence (13%) compared with DE-CMR (20%; p = 0.02). Whereas DE-CMR detected 14 of the 15 thrombi detected by contrast echo (sensitivity 93% based on echo reference), contrast echo identified 14 of 23 thrombi (61%) detected by DE-CMR. Impaired thrombus detection was not specific to contrast echo but was also evident for anatomic imaging by CMR, as shown by the fact that 21% of thrombi were missed by cine-CMR. Figures 1B and 2 provide examples of patients with thrombus detected by DE-CMR but missed by either contrast echo or cine-CMR.
Detection of thrombus by anatomic imaging varied according to location, type, and size. Overall, 75% of thrombi were apical in location and 67% were intracavitary in shape. Apical and nonapical thrombi were similar in size (3.2 ± 4.0 cm3 vs. 5.4 ± 7.4 cm3, p = 0.7). Although apical thrombi missed by contrast echo were, on average, <20% as large as those detected (0.8 ± 0.6 cm3 vs. 4.3 ± 4.5 cm3, p = 0.02), nonapical thrombi missed by contrast echo varied widely in size with thrombi missed (7.3 ± 8.7 cm3, range 0.2 to 19.2 cm3) occasionally larger than those detected (1.5 ± 1.1 cm3, 0.8 to 2.3 cm3) (p = 0.7). As shown in Figure 3, a similar trend was observed for apical thrombi missed versus detected by noncontrast echo (2.0 ± 2.7 cm3 vs. 4.6 ± 5.1 cm3, p = 0.097). Thrombus shape also modified detection by anatomic imaging, with mural thrombus less likely to be detected by contrast echo and cine-CMR (both p < 0.05). Half (3 of 6) of all nonapical thrombi were mural in shape.
This study, the first to simultaneously compare contrast and noncontrast echo with DE-CMR for LV thrombus, provides several new observations: First, among a high-risk population, thrombus detection was markedly improved by a uniform strategy of contrast-enhanced imaging. Patients who derived incremental benefit from DE-CMR or contrast echo had lower LVEF than those in whom noncontrast echo alone accurately assessed thrombus. Second, despite the marked improvement in diagnostic performance yielded by echo contrast, DE-CMR detected more thrombi than did contrast echo. Contrast echo limitations were not modality-specific, as evidenced by the fact that cine-CMR missed 21% of DE-CMR evidenced thrombi, with cine-CMR and contrast echo less likely to detect mural thrombus. Third, infarct size by DE-CMR was an independent marker for thrombus even after controlling for LVEF and aneurysmal dilation.
A major aim of our investigation was to compare an established technique of thrombus tissue characterization by DE-CMR (1,2) with multiple techniques that identify thrombus based on anatomic appearance. Two-thirds of DE-CMR–evidenced thrombi were not detected by noncontrast echo, whereas contrast-echo improved sensitivity nearly 2-fold. Although it is possible that interval resolution of thrombus may have contributed to differences in test performance between echo and CMR, the interval between modalities was short (0.8 ± 1.4 days) and the additive value of contrast-enhanced imaging was shown for echo and CMR, supporting the concept that improved thrombus detection was related to contrast enhancement.
When interpreting these results, it is important to recognize that there was no uniform gold standard for thrombus and that there are inherent limitations to using any given imaging method as a reference standard. However, our study was predicated on prior research that used pathology and clinical data to validate DE-CMR as a highly accurate technique for thrombus (1,2). Consistent with our findings, prior studies have reported that DE-CMR improves thrombus detection versus echo. In a study that included 12 patients with DE-CMR–evidenced thrombus, Mollet et al. (3) reported that echo detected thrombus in 42%. Srichai et al. (2), studying patients with uniform pathology verification of thrombus, reported that sensitivity of echo was 23% versus 88% for CMR. However, both studies may have understated echo capabilities. In the former, comparisons were exclusively made with noncontrast echo. In the latter, CMR with uniform gadolinium use was retrospectively compared with echo with only occasional use of echo contrast.
A central aspect of our protocol involved uniform echo contrast use irrespective of diagnostic findings or image quality of noncontrast echo. Our results show that routine administration of echo contrast and dedicated imaging protocols can improve thrombus detection in at-risk patients. Our findings are especially pertinent in the context of recent product labeling changes concerning risks of ultrasound contrast agents (8) and ensuing debate regarding the risks of echo contrast and the future of widespread use in the context of these labeling changes (12,13). Although prior studies have reported that echo contrast improves LV thrombus detection, use has almost always been reserved for selected patients with suboptimal noncontrast echo quality (4,14). Our results show that although echo contrast improved image quality in the overall population, image quality alone did not predict improved thrombus assessment. For example, among patients with thrombus by DE-CMR, image quality scores did not differ between noncontrast echoes that detected thrombus and those in which contrast use provided additive benefit. This can be partially explained by our observation that thrombus can be present in areas other than the LV apex and thereby missed despite high-quality apical imaging. The marked improvement in thrombus assessment yielded by echo contrast calls into question consensus guidelines that primarily recommend contrast use based on noncontrast echo image quality (5).
Our findings show that LV function can be useful for guiding imaging strategies for thrombus. Noncontrast echoes that were discordant with DE-CMR had lower LVEF than those that were concordant with DE-CMR. Similarly, patients who derived incremental diagnostic utility from echo contrast tended to have lower LVEF than those in whom noncontrast echo alone accurately diagnosed thrombus. LVEF was predictive as measured by either echo or cine-CMR despite variance between the 2 modalities, with differences possibly attributable to the fact that cine-CMR used planimetry whereas echo used linear measurements. Irrespective of differences between modalities, our findings confirm a well-established association between contractile dysfunction and thrombus (1,9), with the incremental utility of contrast-enhanced imaging paralleling higher thrombus prevalence with impaired LVEF. Although DE-CMR provided improved thrombus detection versus both contrast and noncontrast echo, optimization of echo protocols for thrombus remains important for situations in which echo is used to assess ventricular function, or in which DE-CMR is not available or is contraindicated. As our results show improved thrombus detection by DE-CMR, it is reasonable to expect that the advantage of tissue characterization is not modality specific. Perfusion echo can identify thrombus based on tissue characteristics (15), and our findings suggest that this approach may provide added benefit compared with contrast echo solely for LV cavity opacification.
It is important to recognize that this population was at high pre-test risk for thrombus, as reflected by the fact that 20% had thrombus by DE-CMR. Among this high-risk population, transmural infarct size was a marker for thrombus even after controlling for LVEF or aneurysmal dilation. This finding is consistent with our prior study showing that infarct size by DE-CMR was associated with thrombus among a diverse heart failure cohort (1). However, in our prior study, echo predictors were not evaluated. In this study, thrombus was associated with transmural infarct size even after controlling for LVEF and aneurysm measured by either echo or cine-CMR, showing that infarcted myocardium is an independent marker for thrombus irrespective of the modality used to assess LV function/geometry.
Several limitations should be recognized. First, breath-held DE-CMR can be difficult for some patients, and this can affect tolerance of DE-CMR. Free-breathing alternatives, including navigator or single-shot DE-CMR, were not tested. Our protocol was limited to segmented DE-CMR to evaluate one validated method for thrombus tissue characterization and provide a uniform comparison with echo. Second, although our study compared diagnostic performance of anatomic and tissue characterization imaging, the clinical implications of thrombus detected by DE-CMR but missed by echo are not established.
These findings show that DE-CMR is useful for LV thrombus detection among at-risk patients. Thrombus was associated with infarct size, an additional parameter provided by DE-CMR. Although echo diagnostic performance was markedly improved by a strategy of routine contrast use, prevalence was lower than by DE-CMR, especially for mural thrombus and small apical thrombus. Cine-CMR and contrast echo showed similar performance and strong agreement, suggesting that limitations were not modality specific but rather were attributable to intrinsic features of imaging techniques that identify LV thrombus based on anatomic appearance without assessment of tissue properties.
The authors would like to thank Dr. Debra Beneck (Department of Pathology, Weill Cornell Medical College) and Dr. Leonard Y. Lee (Department of Cardiothoracic Surgery, Weill Cornell Medical College) for their valuable assistance with pathology samples and images.
For the full protocol, please see the online version of this article.
Contrast-Enhanced Anatomic Imaging as Compared With Contrast-Enhanced Tissue Characterization for Detection of Left Ventricular Thrombus
This work was supported by a Doris Duke Clinical Scientist Development Award (to Dr. Weinsaft), the Michael J. Wolk Foundation, and Lantheus Medical Imaging (echo contrast manufacturer).
- Abbreviations and Acronyms
- cardiac magnetic resonance
- left ventricular
- left ventricular ejection fraction
- microvascular obstruction
- inversion time
- Received October 17, 2008.
- Revision received March 3, 2009.
- Accepted March 4, 2009.
- American College of Cardiology Foundation
- Weinsaft J.,
- Kim H.,
- Shah D.J.,
- et al.
- Srichai M.B.,
- Junor C.,
- Rodriguez L.L.,
- et al.
- Mollet N.R.,
- Dymarkowski S.,
- Volders W.,
- et al.
- Lang R.M.,
- Biereg M.,
- Devereux R.M.,
- et al.
- ↵New U.S. Food and Drug Administration information on gadolinium-containing contrast agents. http://www.fda.gov/cder/drug/infopage/gcca/default.htm. Accessed June 1, 2007.
- ↵New U.S. Food and Drug Administration prescribing information for Definity approved October 10, 2007. http://www.fda.gov/cder/foi/label/2007/021064s007lbl.pdf. Accessed October 20, 2007.
- Sievers B.,
- Elliott M.D.,
- Hurwitz L.M.,
- et al.
- Main M.L.,
- Goldman J.H.,
- Grayburn P.A.
- Kusnetzky L.L.,
- Khalid A.,
- Khumri T.M.,
- Moe T.G.,
- Jones P.G.,
- Main M.L.
- Kirkpatrick J.N.,
- Wong T.,
- Bednarz J.E.,
- et al.