Background
Longitudinal strain (LS) imaging is an important tool for the quantification of left ventricular function and deformation, but its assessment is challenging in the presence of echocardiographic contrast agents (CAs). The aim of this study was to test the hypothesis that destruction of microbubbles using high mechanical index (MI) could allow the measurement of LS.
Methods
LS was measured using speckle strain (speckle-tracking LS [STLS]) and Velocity Vector Imaging (VVI) before and after CA administration in 30 consecutive patients. Low MI was used for left ventricular opacification and three-dimensional high MI for microbubble destruction. Four different settings were tested over 60 sec: (1) baseline LS without contrast, (2) LS after CA administration with low MI (0.3), (3) LS after CA administration with high MI (0.9), and (4) LS after microbubble destruction with high MI and three-dimensional imaging.
Results
Baseline feasibility of LS assessment (99.3% and 98.2% with STLS and VVI, respectively) was reduced after CA administration using STLS at low (69%, P < .0001) and high (95.4%, P = .0002) MI as well as with VVI (93.8%, P = .004, and 84.7%, P < .0001, respectively). STLS assessment was feasible with high MI after microbubble destruction (1.7% of uninterpretable segments vs 0.7%, P = .26) but not using VVI (7.2% vs 1.8%, P < .001). Regardless of which microbubbles or image settings were used, VVI was associated with significant variability and overestimation of global LS (for low MI, +4.7%, P < .01; for high MI, +3.3%, P < .001; for high MI after microbubble destruction, +1.3%, P = .04).
Conclusions
LS assessment is most feasible without contrast. If a CA is necessary, the calculation of LS is feasible using the speckle-tracking method, if three-dimensional imaging is used as a tool for microbubble destruction 1 min after CA administration.
Echocardiographic contrast agents (CAs) improve endocardial definition, allowing the accurate quantification of left ventricular (LV) volumes and systolic function and the assessment of wall motion abnormalities. The recent use of longitudinal strain (LS) imaging for the quantification of regional and global myocardial function has led to investigation of whether LV opacification with CAs could be combined with LS assessment, adding further information to the quantification of LV mechanics. Unfortunately, the feasibility of LS measurements in the presence of CAs, with either tissue Doppler velocity or Velocity Vector Imaging (VVI; Siemens Healthcare, Erlangen, Germany), was significantly reduced, with substantial interindividual variability, potentially limiting the clinical integration of contrast with VVI. These previous studies had the limitations of (1) small numbers of patients, (2) limited variations in image acquisition settings (only line density), (3) limited assessment of LS in only the apical four-chamber view, (4) lack of comparison between modalities (speckle tracking vs VVI), and (5) the use of a single CA.
Speckle-tracking LS (STLS) has a potentially important role in the identification of subclinical LV dysfunction and has shown incremental prognostic value to ejection fraction, especially when ejection fraction is mildly reduced and in the setting of LV hypertrophy, in which ejection fraction is a less reliable marker of systolic function. In patients with borderline image quality, a desire for accurate volume quantitation, or when masses are sought, CA use may precede the acquisition of STLS and provide technical challenges. We hypothesized that the use of different timing intervals and changes in image settings with higher mechanical index (MI) to promote microbubble destruction could permit the measurement of LS using both STLS and VVI despite the presence of CA.
Methods
Study Design
We prospectively recruited 30 consecutive patients with histories of cancer who were part of a cardiotoxicity surveillance protocol that includes clinically indicated transthoracic echocardiography requiring CA. Two CAs currently approved by the US Food and Drug Administration were tested in the following manner: perflutren lipid-encapsulated microspheres (Definity; Lantheus Medical Imaging, Inc., North Billerica, MA) was used in the first 15 patients, whereas perflutren protein type A microspheres (Optison, GE Healthcare, Princeton, NJ) was used in the subsequent 15 patients. Exclusion criteria were pregnancy or lactation, known allergy to octafluoropropane, presence of right-to-left or bidirectional shunts, and chronic pulmonary vascular disease.
The study protocol was reviewed and approved by the Cleveland Clinic Institutional Review Board, and all patients provided informed signed consent.
Contrast Echocardiographic Protocol
The CAs were used according to the standard protocol described by the manufacturers. A trained registered nurse administered the CAs through an intravenous cannula in a forearm vein on the opposite side of the sonographer’s imaging position. For both CAs, venting was carefully done to ensure that no air was introduced into the vial, potentially destroying the microbubbles.
Definity was agitated for 45 sec in a VialMix activator (Lantheus Medical Imaging, Inc.) before administration. After the mixing was completed, 2 mL of Definity was withdrawn and combined with 8 mL of 0.9% normal saline. A first dose of 1 to 2 mL was administered slowly as an intravenous bolus, with additional incremental doses of 1 to 2 mL administered to optimize LV cavity opacification and avoid far-field attenuation.
For Optison, the vial was inverted, gently rotated for approximately 30 sec to resuspend the microspheres, and then vented before its use, ensuring that no air was introduced, potentially destroying the microbubbles. A 1:1 mixture with 3 mL of saline and 3 mL of the suspension was withdrawn into a 10-mL syringe. An initial bolus of 0.5 to 1.0 mL was given over 10 sec. This allowed adequate LV opacification and avoided attenuation artifacts. Increments of 0.5 mL of Optison (up to 5.0 mL cumulatively in a 10-min period) were given until LV opacification was achieved, while avoiding far-field attenuation.
Adjustments to machine settings, including but not limited to MI and time gain compensation, were manually made by the only sonographer involved in this project (D.A.) to optimize image quality. Contrast recordings were started once optimal LV opacification was achieved.
Echocardiographic Imaging and Contrast Protocol
Echocardiographic studies were performed using Vivid 7 Dimension (SW version 3.1.3; GE Vingmed Ultrasound AS, Horten, Norway) with an M3S probe for two-dimensional imaging, and a single-crystal matrix-array transducer for three-dimensional (3D) imaging (M5S). Grayscale images were obtained at a frame rate of ≥60 frames/sec using harmonic (1.7/3.4 MHz) B-mode imaging. The MI used for routine 2D transthoracic imaging ranged between 0.9 and 1.2, and time gain compensation on the system was optimized to bring out the far field when low-MI contrast was used. Efforts were made to ensure that the plane of interrogation between all sets of images obtained was as similar as possible. LV ejection fraction was calculated offline using semiautomated analysis software (TomTec 4D LV-Analysis version 2.2; TomTec Imaging Systems, Munich, Germany) with contrast-enhanced 3D echocardiography, as this method has been shown to have the best agreement with the gold standard, cardiac magnetic resonance ( r = 0.89).
Two-dimensional LS was assessed using the same 2D probe (M3S) from apical four-chamber, two-chamber, and three-chamber views before and after CA administration, at four different image settings ( Figure 1 ): (1) baseline LS without contrast, (2) LS 1 min after CA administration with low MI (0.3), (3) LS 1 min after CA administration with high MI (0.9–1.2), and (4) LS 1 min after CA administration at high MI (0.9–1.2) using the 3D imaging probe (M5S) for microbubble destruction.
Each of the four different image settings was used for 60 sec to enhance the destruction of contrast microbubbles before acquiring 2D grayscale images for LS measurement. In the last setting, live, real-time 3D imaging was applied for 60 sec in the apical four-chamber view with the objective of covering the entire myocardium. Imaging probes were switched and final image acquisition, for LS measurement, was done with the standard 2D probe (M3S) from apical four-chamber, two-chamber, and three-chamber views.
An 18-segment model of the left ventricle (six segments at each LV level) was used to calculate mean global LS (GLS) for each image setting and for regional LS assessment.
Strain Imaging Measurement
STLS analysis was performed offline using standard software (EchoPAC-PC version 11; GE Healthcare, Milwaukee, WI). Aortic valve closure was used to define end-systole. After manually tracing the endocardial border at end-systole, a region of interest was manually adjusted to include the entire myocardial thickness. Care was taken to avoid including pericardium in the region of interest. Speckles were then tracked frame by frame throughout the entire cardiac cycle.
VVI LS analysis was also performed offline by a single interpreter (J.L.C.) using syngo version 3.0 (Siemens Medical Solutions USA, Inc., Mountain View, CA). For each view, endocardial borders were manually traced in the end-diastolic frame, and the software subsequently traced the borders in the other frames automatically. The vectors of the velocities of the endocardial and epicardial points were then displayed and overlaid onto the B-mode images. Graphical displays of deformation parameters for each segment were then generated automatically and were used for measurement. Similar to STLS, global and segmental LS was measured in the three apical views (four, two, and three chamber).
Statistical Analysis
Data are expressed as mean ± SD. Continuous variables were compared among the four settings using analysis of variance. Mantel-Haenszel χ 2 tests were used for comparisons of dichotomous variables. Bland-Altman plots were constructed to assess and compare the agreement of baseline LS measurements compared with after microbubble destruction (using STLS and VVI) and comparing the two CAs using the latter imaging setting. Statistical analysis was performed using SPSS version 17 (SPSS, Inc., Chicago, IL), and P values < .05 were taken as significant.
Results
Study Group
The overall group consisted of a young (mean age, 52 ± 16 years) and overweight (mean body mass index, 28 ± 5 kg/m 2 ) cohort, with a slightly higher prevalence of women (63%). Hypertension was present in 30% of the patients. Breast cancer was the most common malignancy (47%), and the majority of patients (87%) had undergone prior chemotherapy for their treatment. The baseline characteristics and demographics of the studied group ( Table 1 ) showed no statistical differences in baseline characteristics in the two groups when divided according to the CA used. The mean contrast-enhanced 3D LV ejection fraction was 61 ± 5%. There were no adverse events seen or side effects reported in the studied population.
| Variable | Overall | Definity ( n = 15) | Optison ( n = 15) | P |
|---|---|---|---|---|
| Age (y) | 52 ± 16 | 49 ± 16 | 54 ± 15 | .46 |
| Men | 37% | 33% | 40% | 1.00 |
| Body mass index (kg/m 2 ) | 28 ± 5 | 29 ± 6 | 28 ± 3 | .92 |
| Hypertension | 30% | 33% | 27% | 1.00 |
| Breast cancer | 47% | 47% | 47% | NS |
| Prior chemotherapy | 80% | 73% | 87% | .65 |
| 3D LV ejection fraction | 61 ± 5 | 63 ± 5 | 59 ± 4 | .07 |
Feasibility
The total number of segments used in the analysis for both LS techniques (STLS and VVI) was 540 segments at baseline, 450 at low MI, 414 at high MI, and 450 at high MI after microbubble destruction.
Feasibility was determined by the ability of the automated tracking algorithm to identify the LS of each individual LV segment. As such, contours of the LV endocardial border were drawn at each of the three apical views. Each LV wall was divided into three segments; tracking quality was obtained for each myocardial segment derived with a block-matching algorithm to define the quality of speckle tracking (green = good tracking, red = poor tracking) for each segment. Segments automatically identified by the software that had suboptimal tracking quality after three manual tracing attempts were considered uninterpretable. Figure 2 shows examples of studies with good tracking ( Figures 2 A and 2 C) and others with poor tracking for which segments were excluded using the STLS method ( Figure 2 B). For the VVI method, the adequacy of tracking was verified manually, and the region of interest was readjusted up to three times to achieve optimal tracking ( Figure 2 C). If this was unattainable, that segment was excluded. Poor tracking was assessed subjectively with contours indicating a nonconventional pattern different from the baseline tracing ( Figure 2 D).
Figure 3 shows the percentage of uninterpretable segments according to the LS measuring technique, image settings, and CA. Regardless of which CA used, conventional low-MI settings produced an overall statistically significant worsening for both tracking algorithms, with high percentages of uninterpretable segments ( Figures 3 A and 3 B). At high MI (0.9–1.3), there was an improvement using the speckle-tracking method but significant worsening with the VVI method ( Figure 3 B).
The best setting identified, in the presence of CA, was the speckle-tracking method after 1 min of 3D probe imaging at high MI to cause partial microbubble destruction. These changes did not cause significant worsening in the STLS tracking algorithm (0.7% of uninterpretable segments at baseline vs 2% with high MI after microbubble destruction, P = .10; Figure 4 ). However, the same findings were not seen using the VVI method, for which significant regional and GLS variability and worsening of the tracking algorithm were seen after CA administration, particularly with high-MI settings ( Figure 5 ).
Not all segments were equally affected. For STLS, the most affected were, in decreasing order, the basal and mid anterior and basal inferolateral, regardless of the image setting, except for baseline, for which no segments were affected. For VVI, there was no specific pattern, with the basal and mid inferior and inferolateral segments being the most affected with low-MI imaging, whereas the mid inferior and inferolateral segments were affected using high MI ( Table 2 ).
