Accuracy and feasibility of contrast echocardiography for detection of perfusion defects in routine practice

Comparison with wall motion and Technetium-99m sestamibi single-photon emission computed tomography

Study design

This multicenter study involved acquisition of MCE and single-photon emission computed tomographic (SPECT) data at 11 institutions. Routine clinical practice was mimicked by image acquisition by experienced echocardiographers who had no special expertise in MCE. An independent review of the echocardiographic images was performed by qualitative evaluation of standard gray-scale images by an echocardiographer with experience but not special expertise in MCE. Nuclear images were interpreted independently at the nuclear core laboratory of the Cleveland Clinic Foundation. The institutional review board at each center approved the study, and all patients gave written informed consent.

Patient selection

Two hundred three patients (age 60 ± 11 years, 35 women and 168 men) with MI within the preceding year were recruited to undergo rest technetium-99m sestamibi (MIBI)-SPECT and echocardiography within a 48-h interval. No patients were studied <5 days after MI. Patients were also excluded on the basis of the following criteria: clinically unstable conditions, childbearing potential, hypersensitivity to MIBI and pregnancy or lactation.

The clinical characteristics of the patients are summarized in Table 1. Recent MI (<15 days) was present in 29%, and most patients (78%) had had a single MI. Of the 311 total infarctions, the site was identified by electrocardiography as inferior in 101 (32%) and anterior, septal or apical in 167 (54%); the remainder were lateral or of unknown location. Coronary angiography was performed in 160 patients, with >50% stenoses present in the left main coronary artery in 3%, the left anterior descending coronary artery in 49%, the left circumflex coronary artery in 28% and the right coronary artery in 34%. Previous angioplasty or bypass surgery had been performed in 104 patients, including 56 patients who had one of these procedures within the 3 months preceding the studies.

MIBI-SPECT protocol

Single-photon emission computed tomography myocardial perfusion imaging was performed using standard protocols in all patients (8). Single-photon emission computed tomography imaging was performed 45 to 60 min after injection of 10 to 20 mCi of MIBI. The images were processed on site to check for technical adequacy at the time of the examination, and then sent in digital format to the core laboratory.

Contrast echocardiography

The contrast agent used for this study was a perfluorocarbon-filled microbubble of median diameter 3.4 μm (NC100100; Nycomed Imaging AS, Oslo, Norway). Patients were randomized into seven groups: placebo, three incremental doses studied with fundamental imaging and three doses using harmonic imaging. The low, intermediate and high doses with fundamental imaging were 0.03, 0.10 and 0.30 μl particles/kg body weight of NC100100, and with harmonic imaging they were 0.006, 0.03 and 0.15 μl particles/kg body weight of NC100100. Each patient received four injections of the same dose, separated by at least 5 min, with a 10-ml saline flush to the cannula after each injection.

The images were obtained using standard, commercially available equipment, with harmonic transducers if indicated. Transducer frequency was selected to give the best baseline image, using a medium dynamic range (45 to 55 dB), with maximal power for intermittent imaging and intermediate power for continuous imaging. The focus was set at two-thirds of the depth of the image, and postprocessing was done using the most linear curve. Gain was adjusted to see myocardial tissue speckle on the baseline images, and time and lateral gain compensations were adjusted to give the best homogeneous myocardial brightness on the baseline image. After the first injection of contrast agent, the images were acquired continuously from apical and, if available, parasternal windows. Unless this was impossible for technical reasons, each image set comprised apical four-chamber, apical two-chamber and apical long-axis images, with parasternal long- and short-axis images if available. After continuous imaging, systolic triggered imaging was performed in sequential cardiac cycles gated on the T wave. Separate injections were performed for gated imaging in the apical four-chamber, apical two-chamber and long-axis views. The images were recorded on super VHS videotape and transferred to the core laboratory.

Image interpretation

Echocardiographic images

Videotaped images were digitized using standard equipment (Imagevue; Kodak Health Imaging, Allendale, New Jersey) into a side-by-side display in which rest images were expressed in one quadrant and the other three quadrant were filled by sequential images at various times after the injection. These time intervals varied according to the presence and duration of attenuation and adequacy of myocardial visualization, but usually involved samples at 15, 30, 45, 60, 90, 120 and 180 s after the injection. Interpretations were not made >3 min after the injection.

A single observer interpreted MCE, and reproducibility was analyzed by review of 20% of the studies by a second observer. Overall gain settings and image quality were scored at baseline and during continuous and gated acquisition. Segmental interpretations were made for visibility, gain settings, contrast effect and function.

Myocardial contrast effect was interpreted separately for continuous and triggered image sets, on the basis of a review of both the digital images and videotape (9). Visualized segments were characterized as not opacified, poorly opacified or adequately opacified—the last being identified by opacification of >50% of the segment (Fig. 1). The reviewer interpreted whether inadequate opacification was due to reduced perfusion, attenuation or artifact, uncertain cause or failure to develop a contrast effect. The last entailed failure to develop myocardial contrast opacification because of administration of insufficient contrast agent, and was identified if all segments failed to opacify. When some segments became opacified and others did not, the observer categorized these areas as showing perfusion defect, artifact or attenuation. This distinction was made on the basis of the location, extent and characteristics of the defect. Artifact was most commonly seen in the lateral or anterior walls (Fig. 2), and attenuation was identified if more proximal contrast agent in the wall (Fig. 1)or left ventricular (LV) cavity shadowed the distal part of the image. Segments were classified as not seen if the image quality was inadequate or gain settings were too high or too low to produce visualization of that segment, or if the segment was cut off.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Normal myocardial perfusion by MCE. The initial noncontrast image (A)is of good quality with fairly uniform gain settings. After injection of contrast agent (B),myocardial opacification is seen in the septum, apex and apical lateral wall (small arrows),with an attenuation artifact (large arrow)in the basal lateral wall. The accompanying SPECT image (C)shows normal perfusion by nuclear imaging.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Apparent perfusion defect due to artifact. The noncontrast image (A)shows lower levels of backscatter from the lateral wall, probably due to interposed lung. After injection of contrast agent (B),an apparent defect involves the basal half of the lateral wall (arrows).The accompanying SPECT image (C)shows normal myocardial perfusion. This pattern is the most common source of false positive findings (i.e. reduced MCE when there should be normal perfusion).

The LV function was assessed during a separate reading from the myocardial contrast evaluation. Each segment was classified as showing normal function, hypokinesia, akinesia or dyskinesia (9). In the situations where precontrast images were too poor to make this interpretation, we used cardiac cycles demonstrating LV opacification.

Statistical analysis

Results were analyzed from segmental and patient perspectives. In the segmental analysis, concordance was recorded between the observers for patients in each MCE dose and imaging group, as well as for scintigraphic and wall motion analysis. Segmental sensitivity and specificity were calculated by comparison of the myocardial contrast or wall motion results, or a combination, with segmental SPECT results. Sensitivity was thereby defined as the proportion of moderate and severe SPECT defects identified by MCE as perfusion defects. In contrast, specificity was defined by the proportion of segments with normal SPECT showing normal myocardial contrast opacification by MCE. The limitation of the overall analysis is that a significant number of segments may contribute suboptimal data on the basis of inadequate gain settings or image quality. These situations compromise the accuracy of the echocardiographic technique through mechanisms unrelated to the contrast agent. Thus, subanalyses were performed after exclusion of segments not visualized or with technical limitations such as inappropriate gain settings or inadequate image quality. In each situation the impact of exclusions on test feasibility was expressed. Because of potential bias introduced by interactions between segments, this analysis was repeated using a patient-weighted approach whereby variables were calculated for each patient and then averaged for all patients.

A second patient analysis was performed by defining patients with moderate or severe SPECT defects of more than moderate size as having disease, and those with no or smaller defects as having no disease. If one or more echocardiographic segments showed perfusion defects on MCE, it was identified as a positive MCE response, and if none was identified, as a negative response.

Comparisons between groups were obtained using the chi-square test. Logistic regression models were used to identify independent correlates of correct results. All statistical analysis was performed using SAS statistical software (SAS Software, Cary, North Carolina).

Patient characteristics

The groups randomized to different doses of the contrast agent were comparable with respect to baseline clinical variables (Table 1). Similarly, the groups were comparable with respect to echocardiographic visibility, wall motion score and extent and severity of SPECT defects (Table 2). Among the seven groups, 80% to 89% of the segments were well visualized at baseline. Gain setting and quality were considered adequate in 60% to 81% of the studies, of which 91% to 97% of the segments were characterized as well visualized.

Normal wall motion was present in 63% to 70% of the segments across the dose groups. With respect to the severity of abnormal motion, 13% to 16% of the segments were hypokinetic and 15% to 22% were akinetic. The extent of wall motion abnormalities was most commonly seen in the mild to moderate range (0 to 6 segments), being 76% in the high dose group and 79% in the intermediate dose group. The two observers demonstrated a 79% concordance between their analyses of segmental wall motion.

Most segments (55% to 65%) demonstrated normal myocardial perfusion by SPECT. A concordance of 95% was reported between the two observers in the categorization of perfusion defect severity. Most segments (43% to 68%) showed mild to moderate perfusion defect extent (less than six segments) by SPECT. Similarly, the extent of perfusion abnormality by harmonic MCE was one segment or less in 55% to 64% of patients.

Comparison of MCE and SPECT

Definition of optimal contrast dose and imaging methodology

The concordance between observers, as well as the sensitivity and specificity of MCE for identifying SPECT perfusion defects and normal perfusion using segmental analysis, are summarized in Table 3. For the evaluation of all segments, a trend can be seen supporting the use of triggered imaging in the harmonic mode, and the intermediate and highest doses as demonstrating the most favorable specificity (50% and 67%, respectively).

View this table:

• View inline
• View popup

Table 3

Comparison of Sensitivity and Specificity of Myocardial Contrast Effect for Prediction of Moderate and Severe Single-Photon Emission Computed Tomographic Perfusion Defects at Each Dose and With Each Imaging Technique legend

The first subanalysis involved exclusion of segments contributing suboptimal data on the basis of inadequate gain settings or image quality. After the exclusion of these segments, the sensitivity and specificity were generally increased (Table 3). Again, this analysis showed that the most favorable results were obtained in intermediate and high dose imaging using the triggered harmonic mode. The second subanalysis focused on exclusion of segments where a contrast defect was visualized, but was deemed by the observer to be due to attenuation (Fig. 1), artifact (Fig. 2)or absence of contrast effect. When these negative studies were removed, the number of interpretablesegments (expressed as “feasibility”) decreased further, but exclusion of these segments increased the specificity of the test to 93% using the highest dose with triggered harmonic imaging. The proportion of patients in whom the test was feasible (i.e., delivered data demonstrating either a perfusion defect or normal perfusion) increased with progressively increasing dosage, and was higher with triggered than with continuous imaging. On the basis of these results, Figure 3compares the sensitivity, specificity and accuracy of different imaging approaches, combining intermediate and high dose data. This analysis shows that the most favorable approach was triggered harmonic imaging.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Sensitivity and specificity of intermediate and high dose myocardial contrast agent injection with continuous and triggered harmonic imaging for detection of SPECT defects. Accuracy data reflect segments where interpretation was feasible (proportion of total shown as feasibility).

Table 4summarizes sensitivity, specificity, reproducibility and feasibility using the patient-based approach. This reproduces the same findings as the segmental approach, with the optimal feasibility being in the triggered harmonic mode at the highest dose.

View this table:

• View inline
• View popup

Table 4

Patient-Weighted Data Comparing the Sensitivity and Specificity of Myocardial Contrast Effect for Prediction of Moderate and Severe Single-Photon Emission Computed Tomography Perfusion Defects at Each Dose and With Each Imaging Technique legend

Determinants of MCE with optimal dose and methodology

On the basis of the aforementioned results, subsequent subanalyses focused on harmonic imaging with intermediate and high doses.

The location of the infarct influenced the results of MCE. In patients with MI in the anterior territory, the specificity (79%) of MCE was greater than that in those with MI in the inferior territory (65%). In a logistic regression model, anterior MI was associated with true negative responses (p = 0.03), whereas inferior MI was not (p = 0.63). However, the feasibility of the test in anterior (86%, p = 0.20) and inferior infarctions (75%, p = 0.33) was similar, as was the sensitivity (29% vs. 24%, p = 0.13 vs. 0.41, respectively). Time after MI also influenced the accuracy of the technique. Myocardial contrast echocardiography showed a greater specificity in the setting of recent (5 to 15 days) rather than older infarcts using continuous (52% vs. 27%, p < 0.001) or triggered imaging (82% vs. 66%, p = 0.02).

The relation between the proportion of abnormal segments and the severity of SPECT perfusion defects is illustrated in Figure 4, A. With both continuous and triggered imaging, defects were most commonly associated with severe SPECT defects (activity <60% of maximum). Moderate SPECT defects (between 60% and 75% of peak activity) showed an intermediate frequency of MCE defects, and a low frequency of MCE perfusion defects occurred in the setting of normal SPECT activity (i.e., 80% to 100% of peak).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Relation of SPECT perfusion defect severity with probability of MCE perfusion defect (A)and probability of abnormal wall motion (B). WMA = wall motion abnormality.

Comparison with regional LV function

Normal SPECT perfusion was present in 1,883 segments, 80% of which showed normal regional function. Moderate perfusion defects were reported in 639 segments, among which function was abnormal in 36%. However, 394 (70%) of the 563 segments with severe defects were associated with normal function. The relation between SPECT defect severity and wall motion is described in Figure 4, B.

Because some discrepancies between SPECT and MCE may reflect false positive SPECT defects due to soft-tissue attenuation (in which case wall motion should be normal), the correlation between MCE and SPECT was reanalyzed when SPECT and wall motion were in agreement. This analysis also compensates for discrepancies between SPECT and MCE due to malalignment between the two modalities. Table 5shows that MCE and SPECT were more concordant in segments where SPECT and wall motion were in agreement and patient-weighted data were comparable.

Combination of MCE with LV function data

The simplest means of combining these data is to use wall motion assessment when the MCE data are nondiagnostic. With continuous imaging, this increased the sensitivity and specificity of the intermediate and high dose groups from 20% and 32% to 32% and 87%, also permitting interpretation of all segments. Similarly, the sensitivity and specificity of the intermediate and high dose groups with triggered imaging increased from 24% and 69% to 27% and 92%.

Myocardial contrast echocardiography and assessment of regional LV function were combined using a stepwise approach, whereby wall motion was analyzed first and if this was abnormal, no MCE evaluation was made, on the basis of the relatively high specificity of wall motion abnormalities for SPECT defects. In contrast, if wall motion was normal, segments were evaluated for myocardial contrast defects. The latter were classed as positive for coronary disease, whereas segments showing normal perfusion were classified as negative. This analysis of intermediate and high dose harmonic studies gave a sensitivity and specificity of 55% and 83% for triggered and 46% and 82% for continuous acquisition, respectively.

Detection of extensive perfusion defects

The results were analyzed according to the ability of the test to predict a perfusion defect of significant size in individual patients. Patients were identified as being positive (scar) by SPECT if more than two segments were abnormal, on the basis of prognostic data showing defects of >12% to 15% of the LV as prognostically important. Patients were identified as being positive (scar) by MCE if there were one or more segments with perfusion defects on MCE. The results of this analysis are recorded in Table 6. For this analysis the most favorable results were obtained using the intermediate dose with the triggered harmonic imaging approach.

Influence of technical considerations on MCE

A number of technical details relating to both contrast agents and imaging require consideration during MCE (12). The use of adequate power and appropriate gain settings is critical. Excessive regional backscatter—for example, from scar segments after MI—may prevent recognition of increased backscatter after the injection of contrast agent. The use of inadequate power or gain may preclude recognition of contrast agent, especially at a low dose and when using fundamental imaging. Even in an optimized image, the reflection of ultrasound from different areas of myocardium is frequently heterogeneous (13). Subtraction techniques may compensate for variations in the rest image, permitting more reliable detection of contrast agent and perhaps enhancing the feasibility of the technique (14). Nonetheless, even with reanalysis of the data after exclusion of segments with unsatisfactory gain settings in this study, our results showed a low sensitivity for detection of SPECT perfusion defects.

Although gas-filled agents (including the one used in this study) produce large amounts of reflected ultrasound (15), optimal concentration of contrast agent remains critical. High bubble concentrations can provoke attenuation in the far field, owing to shadowing by the high levels of backscatter imposed by the bubbles proximal to the transducer (15)within the ventricular cavity or more proximal myocardium. In contrast, the density of scan lines in the apex causes more intense destruction of acoustically labile bubbles (16). Thus, opacification of the apex may require a greater concentration of microbubbles than that which is optimal for opacification of other segments. Bolus administration of contrast agent opacifies the apex, whereas the remainder of the myocardium may be shadowed by attenuation. A contrast infusion avoids some of the problems of attenuation, but low doses may lead to erroneous interpretation of an apical defect. The dependence of attenuation on the presence of microbubbles indicates that is unlikely that these issues can be completely addressed by subtraction of the rest image.

Standard echocardiographic imaging planes have some limitations for MCE. Parasternal views are often limited by attenuation caused by right ventricular contrast, so data are mostly derived from the apical views. In these views the most frequent segments identified as nondiagnostic are the lateral segments (13), which typically have the lowest level of backscatter in the rest study (Fig. 2). A number of explanations have been proposed for this variability, and likely causes include the presence of overlying lung and differences in myocardial fiber orientation. In this study the limited feasibility of myocardial contrast due to artifact may reflect the use of standard imaging planes; the use of off-axis images may have more favorably visualized these walls.

Differences between perfusion scintigraphy and MCE

This study was formulated on the basis of a comparison of myocardial contrast activity and scintigraphic perfusion imaging. Although these two techniques are analogous, they have some important differences. First, the localization of tracer is different with each technique. The intensity of the myocardial contrast effect is determined by myocardial blood volume (17), whereas the scintigraphic technique involves cellular uptake of a perfusion tracer (in this case MIBI, which is concentrated within the mitochondria) (18). Thus, whereas the echocardiographic technique examines only the delivery of the contrast agent to the microvasculature, the nuclear technique also involves the need for viable muscle cells.

Other sources of variability arise from the different imaging characteristics of echocardiographic and nuclear techniques. The spatial resolution of these tests are ∼1 mm and ∼1 cm, respectively, causing partial volume effects to have a greater influence on the nuclear rather than the echocardiographic technique. Left ventricular wall thinning or failure to thicken in patients who have had an MI therefore accentuates the intensity of SPECT defects (19). In stunned myocardial segments the presence of a wall motion abnormality may engender an apparent perfusion defect even though perfusion has been restored. Hence, the greatest concordance between SPECT and MCE in this study was obtained in situations where SPECT was in agreement with wall motion analysis (Table 6).

Artifacts may be obtained with either technique, and their causes differ. With perfusion imaging the most frequent cause of artifacts is attenuation due to overlying soft tissue (11). With MCE false positive perfusion defects are most commonly related to attenuation, inadequate gain and anterior and lateral wall artifact (13).

Finally, myocardial perfusion scintigraphy has become established as a highly sensitive technique for the identification of myocardial perfusion, with a reasonable correlation between myocardial blood flow and regional tracer uptake (18). Although similar data using MCE have been gathered from animal models (20), no such relation has been apparent in this study. This may reflect the limitations of MCE in the clinical situation, where chest wall attenuation is an important consideration, as well as in this particular study group, because major reductions of blood flow were uncommon in these patients. Nonetheless, comparison of MCE, SPECT and wall motion findings indicate a better correlation between SPECT and wall motion data than between SPECT and MCE data. Thus, rather than SPECT being too sensitive, the MCE methods used in this study were insensitive for the detection of mild and moderate perfusion defects.

Previous comparisons of MCE and SPECT have not been comparable with those in the present study. In 41 patients imaged after intracoronary injection of sonicated meglumine (21), 51% had normal perfusion using both techniques, and 85% with perfusion defects using both methods had a history of previous MI. The overall concordance between MCE and gated SPECT (78%) exceeded that in this study, reflecting the more reliable delivery of contrast agent using coronary rather than venous administration. However, defect size was also smaller with MCE than with SPECT.

Intravenous MCE and nuclear perfusion imaging have been compared in two previous studies (14,22)of patients with known or suspected coronary disease undergoing dipyridamole stress testing. The presence and location of perfusion defects, as well as their relation to stress, were comparable with MCE and nuclear imaging. These studies have a number of important differences from the current study. First, this study was multicentric, and as such reflects the acquisition and interpretation of the technique in the hands of expert echocardiography laboratories rather than centers specifically dedicated to MCE. Second, our study involved comparison of rest images with rest SPECT in patients who had an MI, which may pose particular problems—for example, partial volume considerations and reactive hyperemia. In contrast, the use of coronary vasodilation may favor myocardial contrast opacification. Third, the analysis in our study involved subjective comparison of unprocessed two-dimensional echocardiographic images. The use of extensive postprocessing, including subtraction of the rest image and color coding (14)or videointensity measurements (22), may enhance the feasibility of the technique.

Study limitations

In addition to the differences of MCE and myocardial perfusion SPECT, comparison of nuclear and echocardiographic techniques accounts for some of the variability seen in this study. The presence of wall motion abnormalities in 20% of segments with normal flow probably reflects the influence of myocardial stunning and subendocardial MI, which may compromise function more so than perfusion. The comparison of nuclear and echocardiographic techniques also presents inherent problems in relation to the coregistration of the segments. Despite attempts to minimize errors caused by comparison of different segments by careful orientation of the longitudinal plane axis on the SPECT images, the finding that many segments with moderately abnormal perfusion demonstrated normal function (Fig. 4, B)is partly due to problems of coregistration. Reanalysis of MCE and SPECT when SPECT and wall motion were in agreement suggests that some of these discrepancies were due to SPECT attenuation and image malalignment.

We sought to separate the functional and MCE components of the echocardiogram by examining them at different times. This separation is difficult, as well as artificial, as MCE is likely to be used clinically by combination with wall motion assessment. However, by using wall motion score when the MCE was nondiagnostic, or allowing the wall motion data to take precedence when this was abnormal, we could improve the accuracy of MCE.

Transient response imaging has been shown to reduce bubble destruction due to ultrasound exposure. At the time of the design of this study, this technique was performed by triggering on every T wave to optimize visualization of myocardium. Subsequently it has become apparent that triggering on every second to fifth cycle further reduces the destruction of bubbles and may augment the density of the myocardial contrast signal (22). Combination of transient response imaging with a contrast infusion may enable quantitation of regional myocardial perfusion. The imaging protocols eventually used for MCE will probably differ from those used in this study; triggered or continuous imaging is unlikely to be used separately, and if bolus injections are performed, they will not necessarily be separated by at least 5 min.

We elected to use qualitative gray-scale analysis of the images to mirror routine clinical practice with current equipment. Subtraction of baseline data, color coding and quantification of contrast agent might have improved the accuracy and feasibility of MCE in this study. However, such postprocessing is not in routine use at present.

Conclusions

The findings of this and other studies indicate that myocardial contrast opacification may be reliably obtained using intravenous injection of second-generation contrast agents in combination with harmonic imaging in the triggered mode. The results of this study reflect the specificity of the test using current equipment and visual interpretation. However, MCE is a complicated technique that requires extensive training, the definition of optimal administration protocols and sophisticated methods of interpretation. Although contrast agents have enhanced the field of MCE, these advances have led to expectations of clinical applications that seem unlikely to be fulfilled on the basis of this study. To fulfill the expectations of MCE as a clinical tool, other developments in image acquisition and postprocessing will be required.