|Home | About | Journals | Submit | Contact Us | Français|
Speckle tracking imaging is a promising new echocardiographic method to assess left ventricular (LV) mechanical dyssynchrony. Our aim was to assess a new speckle tracking regional strain algorithm by comparison to angle corrected tissue Doppler (TD) in an animal model of left bundle branch block (LBBB) and cardiac resynchronization therapy (CRT).
Ten open chest dogs had routine grayscale and tissue Doppler images of the mid-left ventricular (LV) short-axis plane. Electrical activation was altered by pacing from right ventricular (RV), LV free wall, and biventricular sites to create various degree of mechanical dyssynchrony and alter regional function. Segmental time-to-peak strain, peak strain and frame-by-frame strain were measured by angle corrected TD, TD M-mode and by speckle tracking on the same digital cineloop. Of 240 possible paired tissue Doppler and speckle tracking segments, data were available for 222 segments (93%); images with catheter artifacts were prospectively excluded. Comparative overall time-to-peak strain by each method correlated well: r = 0.96, bias=−6 ± 20 ms. Of 80 possible paired M-mode tissue Doppler and speckle tracking segments, strain data were available for 76 segments (95%). Comparative overall time-to-peak strain, peak strain and frame-by-frame strain analysis in 1012 frames by each method correlated well: r=0.98, bias of 1±14ms; r=0.82, bias of 3±7%; and r=0.91, bias of 0±6% respectively.
Regional strain analysis using echocardiographic speckle tracking radial strain strongly correlated with strain by angle-corrected tissue Doppler imaging in an animal model of dyssynchrony. Speckle tracking radial strain has potential for clinical applications.
Echocardiographic strain imaging has been shown to be a useful tool to assess regional left ventricular (LV) function because it can differentiate between active thickening and passive wall motion. Strain imaging has been extensively studied previously using tissue Doppler signal analysis to non-invasively evaluate diseases that affect LV and right ventricular function such as effects of ischemia and mechanical dyssynchrony.1–6 However, a practical limitation of tissue Doppler strain imaging is that measurements are affected by the Doppler angle of incidence. Angle-corrected tissue Doppler strain has been a technicological advance to overcome this limitation; however, data quality remain limited where the Doppler angle on incidence approached 90°.7 A more recent approach to quantify regional myocardial function has been devised from analysis of routine gray-scale two-dimensional echocardiographic images, known as speckle tracking, which can calculate myocardial strain independent of angle of incidence.8–10 A high level of interest continues for quantification of LV dyssynchrony for cardiac resynchronization therapy (CRT).11 In addition, recent data using magnetic resonance imaging have shown that circumferential strain is more sensitive to assess cardiac dyssynchrony than longitudinal motion.12 Our group has also shown that regional radial strain is an important means to quantify ventricular dyssynchrony using tissue Doppler radial strain7, 13 and speckle tracking radial strain.14 Accordingly, our objective was to evaluate a novel color-coded speckle tracking radial strain technique by head to head comparison with angle corrected tissue Doppler strain, in order to quantify dyssynchrony in a pacing animal model of left bundle branch block (LBBB) and CRT.
Ten dogs, weighting 20.9±1.4kg were studied. The protocol was approved by the institutional animal care and use committee and conformed to the position of the American Heart Association on research animal use. All dogs were anesthetized with sodium pentobarbital (30mg/kg induction; 1.0mg/kg−1/h−1 with intermittent boluses, as needed) and mechanically ventilated. Pulmonary artery and LV pressure catheters were placed for continuous hemodynamic monitoring. After a median sternotomy, epicardial pacemaker leads were placed on the right atrium (RA) anterior surface, right ventricular (RV) free wall near the infundibulum, and LV free wall near the mid posterolateral wall to create various degree of dyssynchrony and alter regional function.
All measurements were made with respirations suspended at end expiration to control for the effects of cardiopulmonary interactions. Pacing rates were set 5 to 10 minutes above the intrinsic sinus rhythm and ventricular pacing was performed in atrioventricular delay with 30ms, which was necessary for heart rates typically > 100 min−1. Data were collected during RA pacing as control, RA-RV pacing to induce LBBB, LV free wall pacing, and RA-RV-LV free wall pacing to reflect CRT. The order of pacing protocol steps was randomly varied from animal to animal. Between each ventricular paced rhythm interval, the animals were returned to RA pacing allowing hemodynamic variables to return to baseline levels before the next step in the protocol.
An echocardiographic system with Tissue Doppler capabilities was used (Aplio SSA-770A, Toshiba Medical Systems Corp, Tokyo, Japan) with a 3.0 MHz transducer. Digitized routine and color-coded Tissue Doppler images were acquired from mid-LV short-axis levels using a fixed transducer position. Tissue Doppler system frame rate was 49 Hz with a pulse repetition frequency of 4.5 kHz. Velocity ranges were from 17.0 to ±13.0 cm/s to select the range to maximize the sensitivity.
Color Tissue Doppler data sets were analyzed using offline custom software (AplioQ, Toshiba Medical System Corp). Angle correction was performed as previously described in detail.13, 15–17 Briefly, myocardial vector (V) of motion toward a manually placed point of contraction center was calculated as: V motion = V beam/cosine θ, where θ is the angle of incidence of the ultrasound beam. The sectors where the angle of incidence approached 90° and Doppler calculations are not possible were masked and excluded from analysis. Strain was calculated as time integral of velocity gradient that was calculated along radii of a distance (Δr) toward the LV contractile center. Angle corrected Lagrangian strain (ε) was obtained using displacement values from two pixels with the equation:
where v1, v2 were angle-corrected velocities, t = time and L0 is the original distance between two pixels or difference pitch (3 mm in this study).
Routine B-mode gray scales images were analyzed by novel software (Toshiba Corp.) for frame-by-frame movement of stable patterns of natural acoustic markers, or speckles, present in ultrasound tissue images over the cardiac cycle. A circular region of interest was traced on the endocardial and epicardial border of the mid-LV short axis view, using a point-and-click approach. Speckles within the region of interest were tracked in subsequent frames. The location shift of these speckles from frame to frame, which represents tissue movement, provides the spatial and temporal data. This tracking determines the motion of the endocardial and epicardial border tracings through the cardiac cycle. Radial strain was calculated as change in length/initial length between endocardial and epicardial trace. Accordingly, myocardial thickening was represented as positive strain, color-coded as yellow, and thinning was represented as negative strain, color-coded as blue, and these were then superimposed on the conventional 2-dimensional image. The software then automatically divided the short-axis image into 6 standard segments. The region of interest was fine-tuned by visual assessment during the cineloop play feature to ensure that all wall regions were included throughout the cardiac cycle. Often, the regions of interest were placed slightly wider than the LV wall to obtain the highest quality time-strain curves.
Radial strain was measured by angle corrected tissue Doppler radial strain imaging and speckle tracking radial strain on the same digital cine loop using 6 similar regions of interest during 4 different pacing modes (RA control, RA-RV, RA-LV, and RA-RV-LV) in 10 dogs. Regions of interest were manually drawn on 6 segments of the mid-LV short-axis view (anteroseptum, anterior, anterolateral, posterolateral, posterior, and inferior) (Figure 1). Linear polygons with transmural length ranging from subendocardium to subepicardium were placed at end diastole. A tracking algorithm was used with manual adjustment of the size and shape of the regions of interest to maintain tracking of subendocardium and subepicardium throughout the cardiac cycle frame-by-frame. Time-strain curves were constructed, and peak strain and time-to-peak strain from the onset of the QRS were analyzed in all 6 segments.
As this version of speckle tracking software recognized Audio Video Interleave (AVI) formatted file, the same digital cine loop used for tissue Doppler strain analysis was converted to AVI format as a gray scale image without any compression. Care was taken to select a uniform point on the electrocardiogram and confirm the R-R interval was the same as the image used for tissue Doppler analysis when selecting one beat for speckle tracking. After running the speckle tracking software, the widths of regions of interest were manually adjusted to fit with the regions of interest during the analysis of tissue Doppler strain. Both tracking profiles of the region of interests were analyzed with 2 different software packages in an un-blinded fashion. Slight adjustments were made on a frame-by-frame basis for the placement of the region of interest as required by movement artifact. The peak strain value and time-to-peak strain from the onset of the QRS were also analyzed in all 6 segments using speckle tracking strain.. Time-to-peak strain and peak strain were directly compared using 2 different techniques in individual 6 segments. Radial dyssynchrony was defined as the maximum difference in time from earliest to latest segment among 6 segments, and compared by 2 different techniques during different pacing modes.
A complementary comparison study was done with a different method of calculating radial strain using angle corrected tissue Doppler M-mode strain during 2 different pacing modes (RA control, RA-RV) in the same 10 dogs. Angle correction of tissue Doppler velocity image was performed in a similar fashion, and then four M-mode lines were drawn on the mid-LV short-axis view (anteroseptum, anterior, posterolateral, and inferior) (Figure 2). Each M-mode sample comprised averaged data analyzed over 5 adjacent lines with 2.0mm width. After the M-mode data set was reconstructed, two regions of interest were selected at subendocardial and subepicardial layers on this angle corrected Tissue Doppler M-mode velocity screen, and were tracked automatically based on the tissue velocity.18, 19 Tracking of the regions of interest were visually confirmed on both 2-dimensional screen and M-mode screen.
The region of interest was fine-tuned by visual assessment when tracking was insufficient. The tracking lines of both subendocardium and subepicardium were exported as text file with the spatial and temporal data. Radial strain was calculated as change in length/initial length between endocardial and epicardial trace on the Excel spreadsheet. Speckle tracking radial strain analysis was performed in the same way as method of part 1, and 4 regions of interest were manually adjusted to fit with the regions of interest during the analysis of tissue Doppler M-mode strain. Time-to-peak strain, peak strain, and frame-by-frame strain measured from QRS onset to peak strain were directly compared using 2 different techniques in individual 4 segments.
All group data were presented as mean ± standard deviation. Differences were analyzed using a two-way repeated ANOVA approach with the 6 segments and the 4 pacing modes. Correlation analysis were performed with 95% confidence intervals calculated by Fisher r to z transformation, and Bland-Altman analysis.20 The threshold of statistical significance was p <0.05.
Of 240 possible segments comprising 6 segments in 10 dogs during 4 different pacing modes, paired tissue Doppler radial strain and speckle tracking radial strain were available for 222 segments (93%) (Figures 3,,4).4). Segments with catheter artifacts, usually the inferior septal and/or inferior segments, were prospectively excluded. Time-to-peak strain overall by each method correlated well with r=0.96, p<0.001. Bland-Altman analysis revealed no significant bias, with a mean of −6ms and 40ms of limits of agreement (Figure 5). Peak strain values overall by each method were less closely correlated with r=0.51, p<0.001. Accordingly segmental time to peak strain information was more closely comparable than the absolute peak strain values. Bland-Altman analysis revealed no significant bias, with a mean of 0% and 22% of limits of agreement. Fractional area change and radial dyssynchrony were compared during 3 different pacing modes (RA, RV, and biventricular) (Table 1). Radial dyssynchrony was minimal during RA pacing control. Dyssynchrony was increased significantly during RV pacing, stimulating LBBB, while fractional area change diminished. Both LV dyssynchrony and fractional area change improved with biventricular pacing.
Of 80 possible segments comprising 4 segments in 10 dogs during 2 different pacing modes, paired tissue Doppler M-mode strain and speckle tracking radial strain were available for 76 segments (95%). Segments with catheter artifacts were prospectively excluded. Comparative time-to peak strain overall by each method correlated well with r=0.98, p<0.001, with a mean bias of 1 ms and 28 ms of limits of agreement (Figure 6). Comparative peak strain overall by each method correlated well with r=0.82, p<0.001, with a mean bias of 3% and 14% of limits of agreement. Comparative frame-by-frame strain from QRS onset to peak strain by each method correlated well with r=0.91; 95% confidence interval 0.90 to 0.92, with a mean bias of 0% and 12% of limits of agreement (n=1012 frames, Figure 7). Bland-Altman analysis for frame-by-frame strain comparison demonstrated a tendency for speckle tracking to underestimate strain amplitude. The low bias and tight limits of agreement for comparison of negative strain versus positive strain separately showed a mean bias of −3% and 10% of limits of agreement for negative strain, and a mean bias of 1% and 11% of limits of agreement for positive strain, suggesting reasonable agreement between the two methods. Furthermore, overall, time-to-peak strain, strain value, and frame-by-frame strain compared with 2 different techniques demonstrated good correlations for all parameters after eliminating catheter induced artifacts. We confirmed that time to peak strain measured by tissue Doppler strain was correlated with time-to-peak by speckle tracking regardless of pacing modes.
Intra- and inter-observer variability was analyzed in 10 randomly selected studies for segmental time-to-peak strain and global peak strain by speckle tracking analysis using standard 6 segments. Intra- and inter-observer variability for segmental time-to-peak strain expressed as the mean percent error (absolute difference/mean) was 4 ± 8% and 4 ± 8%, respectively (n=56 segments), and for global peak strain was 13 ± 9% and 15 ± 9%, respectively (n=10 studies).
This study demonstrated that assessment of radial strain for dyssynchrony analysis using a new echocardiographic speckle tracking software program produced similar results as angle corrected tissue Doppler in an animal pacing model. Specifically, speckle tracking quantitative analysis yielded similar regional timing as both two-dimensional angle-corrected tissue Doppler and M-mode tissue Doppler imaging techniques for several pacing modes of dyssynchrony and CRT. This study extends the previous clinical experience using speckle tracking dyssynchrony analysis by adding direct comparison to tissue Doppler in a rigorously controlled animal model, which is not possible in humans. This appears to be clinically practical because speckle tracking analysis may be applied to routine two-dimensional gray scale images and it is not affected by ultrasound beam angle of incidence, like tissue Doppler methods. These experiments support the potential for applications of speckle tracking radial strain to quantify dyssynchrony and effects of CRT in humans.
We have previously demonstrated the potential utility of using speckle tracking radial strain to quantify dyssynchrony and predict response to CRT in heart failure patients.14 In particular, differences in measures of time to peak radial strain from the anteroseptal to posterior wall segments ≥ 130 ms, appeared to offer the best balance of sensitivity and specificity to predict response to CRT. More recently, a combined approach which utilized both tissue Doppler for measurement of longitudinal velocity and speckle tracking for measurement of radial strain has been shown to be of additive value to either technique alone.21 Interest remains high in echocardiographic quantification of mechanical dyssynchrony, despite recent reports of potential difficulties.22 An important subset of patients with a wide QRS complex may not have mechanical dyssynchrony, and not respond to CRT. Accordingly, identifying these patients with an imaging technique may potentially improve patient selection for the future.
Dyssynchrony analysis using echocardiography has developed mainly utilizing analysis of longitudinal dynamics imaged from the apical transducer position. Recent work using magnetic resonance imaging has demonstrated that circumferential myocardial dynamics may characterize LV dyssynchrony in a more sensitive manner than longitudinal.12 Therefore, in addition to longitudinal dynamics, radial / circumferential analysis provides important insight into the physiology and mechanics in LBBB and CRT. Previous reports for assessing radial dyssynchrony used gray scale M-mode23, 24 and tissue Doppler radial strain.7 M-mode analysis is influenced by translational effect, and its low reproducibility has been reported.22, 25 Radial strain using speckle tracking can be analyzed among the entire LV segments of short axis image and has a potential for clinical application as we reported previously.14 Although this current study used one vendor’s software, preliminary studies from our laboratory using other ultrasound systems demonstrate similar results for evaluating dyssynchrony by radial strain analysis (Abstract: Tanaka H, Hara H, Saba S, Gorcsan J. Comparison of dyssynchrony using radial strain analysis by different vendors to predict response to cardiac resynchronization therapy. J Am Soc Echocardiogr 2008;21:520.)
The main advantage of speckle tracking is that it is independent of angle of incidence. From the point of view of angles of echocardiography, this benefit will impact mostly on short axis analysis. Longitudinal strain analysis using tissue Doppler can be analyzed in the majority of segments except apex, because of an unfavorable Doppler angle of incidence. Angle corrected tissue Doppler technique was developed to overcome this disadvantage for short axis view, however 20–30% of LV from short axis views cannot be analyzed because Doppler calculation was impossible in the segments with angles of incidence approached 90°.
A limitation of this animal model was that the epicardial placement of leads is not identically comparable to humans with heart failure and the coronary sinus transveous approach. However, similar therapeutic results of biventricular pacing may be achieved with a surgical epicardial approach in humans. Furthermore, a limitation is that this animal pacing model with acute alterations in ventricular contraction sequence may not reflect the precise abnormalities seen in chronic human heart failure with native conduction system disease. However, our group and others have successfully used this model of right ventricular outflow tract pacing to induce abnormal mechanical activation that simulates LBBB.12, 13 Further study in heart failure patients is required. Speckle tracking analysis is dependent on frame rate. Frame rates that are too low often results in greater frame to frame change in the speckle patterns, impacting the performance of the speckle tracking software, and preventing the precise characterization of regional myocardial motion. In this study we used fixed frame rate (49 Hz) for all images. If we performed the analysis using a faster frame rate, discrimination of myocardial motion and deformation may be better improved.14 Image quality of gray scale image was also important factor for speckle tracking strain analysis. In the current study, several catheters were in the ventricular cavities during study. Segments under artifacts lost speckle, and caused unreliable strain data. These artifacts mainly affected speckle tracking analysis as compared with tissue Doppler analysis. Therefore, we prospectively excluded these segments from the analysis. Accordingly, application of speckle tracking is limited in human subjects with in-dwelling catheters or pacing wires which project shadows on the myocardium. Another limitation of speckle tracking is the user interface required for endocardial and epicardial border tracing. These 2 tracing lines and their tracking directly impact on strain data. When tracking failure was observed, fine tuning of tracing, changing the frame for tracing, and increasing gain were performed to improve tracking. Accordingly, placement and adjustment of the regions of interest for both speckle tracking and tissue Doppler were operator dependent and required training and experience. An additional limitation is that open chest dogs have whole heart translational motion. This was minimized with a pericardial cradle, and was further minimized by using angle-corrected tissue Doppler strain imaging and speckle tracking strain imaging, which assess active wall thickening, rather than passive motion.
In conclusion, dyssynchrony analysis using speckle tracking radial strain correlated well with those by angle-corrected tissue Doppler strain imaging during control, pacing induced dyssynchrony, and CRT. This echocardiographic method has promise for applications to patients undergoing CRT.
This study was supported in part by NIH awards HL04503, HL067181 and HL073198.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Disclosure: Toshiba Medical Systems, Corporation provided the ultrasound equipment and analysis software used in this study.