CLINICAL/ORIGINAL PAPERS. Brage H. Amundsen 1 *, Jonas Crosby 1, Per Arvid Steen 2, Hans Torp 1, Stig A. Slørdahl 1,3, and Asbjørn Støylen 1,3

Transcription

1 European Journal of Echocardiography (2009) 10, doi: /ejechocard/jen201 CLINICAL/ORIGINAL PAPERS Regional myocardial long-axis strain and strain rate measured by different tissue Doppler and speckle tracking echocardiography methods: a comparison with tagged magnetic resonance imaging Brage H. Amundsen 1 *, Jonas Crosby 1, Per Arvid Steen 2, Hans Torp 1, Stig A. Slørdahl 1,3, and Asbjørn Støylen 1,3 1 Department of Circulation and Medical Imaging, Norwegian University of Science and Technology, MTFS, N-7489 Trondheim, Norway; 2 Department of Medical Imaging, St Olavs Hospital/Trondheim University Hospital, Trondheim, Norway; and 3 Department of Cardiology, St Olavs Hospital/Trondheim University Hospital, Trondheim, Norway Received 4 March 2008; accepted after revision 25 June 2008; online publish-ahead-of-print 23 July 2008 KEYWORDS Echocardiography; Method comparison; MRI; Strain; Speckle tracking; Tissue Doppler Aims Compare four different echocardiographic methods, based on tissue Doppler imaging (TDI) and speckle tracking (ST) separately or combined, for long-axis strain and strain rate (SR) measurements, using magnetic resonance imaging (MRI) tagging as a reference. Methods and results In 21 subjects (10 with myocardial infarction) peak systolic strain and systolic and early diastolic SR were measured by four different echo methods: (i) two-dimensional (2D) strain (B-mode); (ii) ST (custom software) of segment end-points (B-mode); (iii) similar to (ii), but combining ST with tissue Doppler tracking; (iv) strain from tissue Doppler velocity gradients (VG). Agreement with MRI tagging was better for strain than for SR. Ninety-five per cent limits of agreement were wider for the TDI-VG method, and 2D strain showed negative bias compared with MRI tagging and the other echo methods. Reproducibility was better for 2D strain than for MRI tagging and the other echo methods. Conclusion ST alone or combined with TDI seems to be suitable for automated measurements of regional myocardial deformation. The study gives important information on the strengths and weaknesses of the different methods, which is important for further development to increase accuracy and applicability. Introduction Regional myocardial strain and strain rate (SR) in the left ventricle (LV) can be measured by velocity gradients (VG) from tissue Doppler imaging (TDI). 1 SR is related to contractility, while strain is more closely linked to stroke volume. 2 SR can help to predict functional recovery after revascularization, 3 and give prognostic information in dobutamine stress echocardiography. 4 However, the clinical use of SR and strain measured by tissue Doppler is limited to experienced users due to the low signal-to-noise ratio. 5 For those with limited experience or limited time to do manual analyses, a more automated approach would be advantageous. 6 An automated method would also provide more objective measurements. * Corresponding author. Tel: þ (work)/þ (home); fax: þ address: brage.h.amundsen@ntnu.no Recently, improved hardware and software have allowed angle-independent quantification of myocardial deformation based on speckle tracking (ST) methods in B-mode or radiofrequency (RF)-data Till now, only one study has compared long-axis strain measurements by both TDI and ST with magnetic resonance imaging (MRI) tagging. 11 No study has compared measurements of SR. The purpose of the present study was to compare regional myocardial long-axis strain and SR measured by four different automated echo methods, based on tissue Doppler, ST or a combination, with measurements by MRI tagging. Methods Study sample Twelve patients with recent myocardial infarction (.3 weeks ago) and 11 healthy subjects were included after having given written Published on behalf of the European Society of Cardiology. All rights reserved. & The Author For permissions please journals.permissions@oxfordjournals.org.

2 230 B.H. Amundsen et al. Table 1 Subject characteristics [values are median (range)] Healthy subjects (n ¼ 11) Patients (n ¼ 10) Age (years) 22 (20 32) 61 (50 71) Female/male 6/5 1/9 End-diastolic volume 113 (84 164) 133 (80 218) (ml) Ejection fraction (%) 53 (47 64) 41 (19 58) informed consent (Table 1). Two patients withdrew during the MRI examination due to claustrophobia. No selection or exclusion was done based on image quality, and the time between the echo and MRI examinations was,3 h in all but two subjects (,24 h). The study was approved by the Regional Committee for Medical Research Ethics, and conducted according to the Declaration of Helsinki. Echocardiography The subjects were examined in supine position, with a Vivid 7 scanner and a 2.0 MHz M3S probe (GE Vingmed Ultrasound, Horten, Norway). The three standard apical views were acquired (four-, two-chamber and apical long-axis) with B-mode (second harmonic full-sector single-focus images, frame rate frames/s) and TDI (TDI frame rate frames/s, TDI vs. B-mode frame rate ¼ 3:1, default radial averaging). Analysis A single observer analysed all images. Blinding was ensured by assigning different ID-numbers to the echo and MR-recordings, and by analysing the echo recordings in different order with each method. Strain and SR were analysed in six segments in each view, by four different echo methods (Figure 1): (i) two-dimensional (2D) strain in B-mode images (EchoPac PC BT06, GE Vingmed Ultrasound). 9 The average segmental values were used; (ii) ST of seven points (ST-7P). In-house ST (sum of absolute differences, SAD) software (GcMat) written in Matlab (MathWorks, Natick, MA, USA), where seven kernels (5 5mm 2 ) tracked the displacement of the segment borders in B-mode images and calculated strain and SR from the change of length of each segment; 7 (iii) combined TDI and ST (TDI&ST). This method was similar to ST-7P, but used tissue Doppler data to track the motion of the seven kernels in the beam direction and ST in the underlying B-mode data in the lateral direction. This made the measurements angle independent. Tissue Doppler data between each B-mode frame were averaged, resulting in temporal smoothing; 6 (iv) tissue Doppler-VG (TDI-VG). This method calculated SR and strain from the tissue Doppler VG in a region of interest (ROI) positioned between the segment end-points, which were tracked by TDI&ST. Thus, the ROI followed the same myocardium through the cardiac cycle. Strain and SR were calculated over a distance of mm, and 10 ms temporal averaging was used. In TDI-VG, ST was only used for ROI-tracking, and not for calculation of strain or SR. No spatial smoothing across segments was applied in ST-7P, TDI&ST, and TDI-VG. In ST-7P, TDI&ST and TDI-VG aortic valve closure (AVC) was automatically detected in the mitral annular velocity curve. 12 Peak systolic strain and peak systolic (S SR ) and early diastolic (E SR ) SR were automatically extracted from the curves. 6 The same event in the mitral annular velocity curve was used for timing of AVC in 2D strain. Segments were excluded after visual inspection of the tracking at reduced playback speed, or due to noise in the curves. The tracking score in 2D strain was overruled on the same basis. In TDI-VG segments were excluded if the angle between the wall and the beam was.308. Strain was calculated as Lagrangian strain, while SR was calculated as Eulerian SR. The necessary corrections were applied in the methods, which derived SR from strain. Figure 1 Schematic illustration of the different analysis methods. ST-7P, speckle tracking of seven points (B-mode images); TDI&ST, combined tissue Doppler imaging and speckle tracking; TDI-VG, strain rate (SR)/strain from tissue Doppler velocity gradients. In TDI-VG speckle tracking was only used to track the motion of the ROI, not to calculate strain or SR. Evaluation of tracking quality in speckle tracking of seven points To evaluate the potential of different markers of tracking quality, the cross-correlation coefficients between each kernel and its best SAD-match in the next frame was stored. This coefficient was not used for tracking. For the same purpose, we also stored the spatial distance between the two kernel positions from the forward and backward tracking, averaged over the entire cycle. The apical kernel was excluded from these analyses due to the frequent presence of near-field reverberations. Magnetic resonance imaging tagging MRI tagging was performed on a Philips Intera 3Tscanner with a dualquasar 40/80 mt/m gradient system and a six-channel SENSE cardiac coil. Two separate acquisitions with parallel tagged lines orthogonal to each other were acquired by complementary spatial modulation of magnetization using a multi-phase ECG-triggered T 1 -Fast Field Echo sequence 13 with the following settings: M2D, multi-shot echo planar imaging (EPI) factor 11, echo time (TE)-shortest, flip-angle 208, 24 ms between frames, field-of-view 330 mm, matrix , tag spacing 8 mm, slice thickness 8 mm. As with echo, the four-, two-chamber and apical long-axis views were acquired. Images were analysed in TagTrack (Gyrotools Ltd, Zurich, Switzerland) using a peak-combination harmonic phase algorithm. 13 The tracked contour was exported to Matlab, where seven points along the

3 Regional myocardial long-axis strain and SR measurements 231 contour, corresponding to the segment boundaries, were selected, and segmental strain and SR calculated as in ST-7P and TDI&ST. Temporal smoothing (average of two consecutive samples) was applied before the derivation of SR from the strain curves. Timing of AVC in MRI tagging recordings was guided by the AVC-value found in the tissue Doppler images, and by inspection of the strain curves obtained by MRI tagging, where a notch could be seen at AVC in some segments. Reproducibility for all methods was analysed as: (i) intra-observer (same cycle); (ii) intra-observer (different cycle; two separate acquisitions in MRI tagging); (iii) inter-observer [same cycle as in (i)]. Statistics All values are reported as mean + SD, and a two-sided P, 0.05 was considered a marker of statistical significance. The proportions of analysed segments by each method were compared using Cochran s Q-statistics. The agreement between strains measured by the different methods was evaluated by Bland Altman statistics with calculation of the 95% limits of agreement. 14 For SR there was a significant relationship between the average value and the absolute difference between methods for all comparisons, meaning that the absolute difference between the methods was larger for segments with more deformation. This is a violation to one of the assumptions in conventional Bland Altman analysis, and we therefore chose to compare SR measurements by using correlation coefficients and scatter plots with the line of identity. Paired-samples t-tests were used to compare the different echo methods with MRI to make sure that only segments with measurements by both methods were included. One-way ANOVA was used to compare values at different levels of the LV in the healthy subjects. No post-hoc correction of the P-value for comparison between the groups was applied, but statistically significant tests were only reported if the overall analysis of variance (ANOVA) test had a P, Reproducibility for both strain and SR measurements was calculated as the coefficient of variation (COV). Analyses were made in SPSS 12 (SPSS Inc., Chicago, IL, USA). Results Feasibility The percentage of analysed segments was: MRI tagging: 83%, 2D strain: 81%, ST-7P: 80%, TDI&ST: 83%, and TDI-VG 64% (P, 0.001). In TDI-VG 14 percentage points of the exclusions were due to angle misalignment. Agreement Heart rate was similar during the echo and the MRI examinations ( vs bpm, respectively, P ¼ 0.18). Peak systolic strain The agreement between MRI tagging and the echo methods is shown in Figure 2. Agreement between the echo methods is shown in Figure 3. The 95% limits of agreement vs. MRI tagging were wider for the TDI-VG method than the other echo methods (P, 0.05). 2D strain had a significant negative bias compared with MRI tagging and the other echo methods (P, 0.05). The 95% limits of agreement were similar for peak systolic and peak overall strain (data not Figure 2 Agreement between magnetic resonance imaging (MRI) tagging and the different echo methods for peak systolic strain. First row: Bland Altman plots of the mean and differences (MRI tagging 2 echo) between the methods, with the 95% limits of agreement indicated (dashed lines). Second row: scatter plots with line of identity for each of the four echo methods (abscissa) vs. MRI tagging (ordinate) in patients. All correlations had P, Third row: same plots for controls. *Significant bias, P, Inferior agreement vs. other methods, P, 0.05.

4 232 B.H. Amundsen et al. Figure 3 Agreement between the echo methods for peak systolic strain. Upper right panels: agreement between the different echo methods, shown as Bland Altman plots, with 95% limits of agreement indicated (dashed lines). The difference on the abscissa is calculated as (row method 2 column method). *Significant bias, P, Lower left panels: scatter plots for patients only, with lines of identity. Ordinate axis: column method, abscissa: row method. All correlations had P, shown). The width of the 95% limits of agreement intervals for the echo methods vs. MRI tagging was similar for patients and healthy subjects. The agreement between the methods at different levels of the LV was assessed in the healthy subjects only, as the variable location of regional dysfunction would influence the results in patients (Table 2A). Peak systolic and early diastolic strain rate SR measurements from MRI and the echo methods are compared in Figure 4. 2D strain and the combined TDI&ST method correlated better with MRI tagging than the other methods for both S SR and E SR. Comparison between the echo methods is shown in Figure 5. Overall, MRI tagging measured higher absolute values for S SR, but not E SR, compared with the echo methods (Figure 4). TDI-VG measured higher E SR than the other echo methods. Values for the different levels of the LV are shown in Table 2B and C. Reproducibility The COV for all methods are shown in Figure 6. For the strain measurements, only MRI tagging showed a significant relationship between the mean and difference between the repeated measurements. For SR such a relationship was present for all methods. Automated quality assessment of speckle tracking The mean of the forward and backward cross-correlation coefficients in segments excluded due to dropouts (n ¼ 31) or reverberations (n ¼ 18) were lower than in included (n ¼ 327) segments [median (range): 0.51 ( ) vs ( ) vs ( ), respectively, P, 0.02]. The difference between the forward and backward tracking positions was highest for segments excluded due to dropouts [3.4 ( ) mm], followed by included segments [1.9 ( ) mm], and segments excluded due to reverberations [1.3 ( ) mm, P, 0.01 for dropout vs. reverberation and included, P ¼ 0.07 for reverberation vs. included]. The tracking-quality tool in 2D strain suggested exclusion of 65 of 378 segments, and 40 (62%) of these were agreed upon by visual inspection. Twenty-nine segments were excluded by visual inspection, but not by the trackingquality tool. This gave an overall prospective accuracy of 86%.

5 Regional myocardial long-axis strain and SR measurements 233 Table 2 subjects Peak systolic strain, peak systolic strain rate (SR), and peak early diastolic SR at different left ventricular levels in the healthy MRI tagging 2D strain ST-7P TDI&ST TDI-VG Mean SD Mean SD Mean SD Mean SD Mean SD (A) Peak systolic strain at different LV levels in the healthy subjects Basal Mid-ventricular * * * 6 Apical P (ANOVA) (B) S SR at different LV levels in the healthy subjects Basal Mid-ventricular * Apical P (ANOVA) (C) E SR at different LV levels in the healthy subjects Basal Mid-ventricular 1.7* Apical P (ANOVA),0.001, ,0.001 P-values refer to comparison between left ventricular (LV) levels within each method (comparison within columns): *P, 0.05 basal vs. mid-ventricular; P, 0.05 mid-ventricular vs. apical; P, 0.05 apical vs. basal. Figure 4 Comparison between magnetic resonance imaging (MRI) tagging and the echo methods for systolic and early diastolic strain rate (SR). Upper row: scatter plots with line of identity for comparison of peak systolic SR between MRI tagging (ordinate axis) and the echo methods (abscissa). The correlation coefficients in parentheses are for patients. Lower row: corresponding figures for peak early diastolic SR. *Significant correlation, P, Discussion The present study shows that ST methods have potential for automated analysis of regional myocardial strain and SR. It also shows the potential of methods combining tissue Doppler and ST, and further possible combinations of the methods should be explored. Differences between the B-mode methods Both 2D strain and ST-7P use only B-mode information for myocardial tracking. 2D strain includes spline interpolation, while ST-7P does not. Thus, the results obtained by ST-7P are more representative of the basic ST principle than the application 2D strain. A further difference between the methods is that 2D strain tracks several small kernels distributed along and across the wall, while ST-7P tracks seven larger mid-wall kernels. Thus, spatial resolution in ST-7P is equal to segment length, while the spatial resolution in 2D strain is likely to be larger than one segment due to interpolation along the continuous ROI. Thus, for ST-7P, robustness is achieved by sacrificing spatial resolution instead of applying spatial smoothing.

6 234 B.H. Amundsen et al. Figure 5 Comparison between the echo methods of peak systolic and early diastolic strain rate (SR). Lower left panels: scatter plots with line of identity for comparison of systolic SR between the echo methods. Ordinate axis: column method, abscissa: row method. The overall correlation coefficient is given below each plot, with the coefficient for patients in parentheses. Upper right panels: same plots for early diastolic SR. Note the different scale. All correlations had P, Finally, the ROI in 2D strain is curved, while in the other echo methods and in MRI tagging, strain and SR are measured over segments defined as straight lines. The size of the difference will depend on the curvature and the amount of wall thickening in the segment. Feasibility TDI-VG had lower feasibility than the other methods, mostly due to the method-specific angle-dependency criterion. A previous stress echocardiography study also found that TDI-VG had lower feasibility than TDI&ST, but this did not lead to a different diagnostic accuracy for the two methods. 15 The applied spatial smoothing in 2D strain did not improve feasibility compared with ST-7P or TDI&ST. In MRI tagging most excluded segments were in the basal inferior or infero-lateral wall, where susceptibility artefacts frequently occur. Agreement Peak systolic strain The inferior agreement found for TDI-VG might have been caused by angle deviations,308. Further, quality assessment was easier in the other methods, where the kernel tracking could be visually inspected in low playback speed and compared with the actual motion of the myocardium. Again, as the previous diagnostic study showed a similar diagnostic accuracy between TDI-VG and TDI&ST, the inferior agreement with MRI may be of less clinical importance. 15 The 2D strain method measured more negative strain values than MRI tagging and the other echo methods. This has also been reported previously, and a new version of the software has been tested in vitro, but not in vivo. 16 ST-7P measured somewhat less negative values than 2D strain, especially in the more curved basal and apical segments, suggesting that the bias was an effect of the

7 Regional myocardial long-axis strain and SR measurements 235 Accurate timing of end-systole is essential for separation of systolic and post-systolic shortening. The fact that the 95% limits of agreement were very similar for peak systolic and peak overall strain suggests that the approach for defining AVC in the MRI tagging data was reasonably accurate. Figure 6 Reproducibility of the different methods. Bars illustrate the coefficient of variation for the different aspects of reproducibility in 10 randomly selected subjects (five with myocardial infarction): (A) peak systolic strain; (B) peak systolic strain rate (SR); (C) peak early diastolic SR. curvature and thickness of the 2D strain ROI. Alternatively, only 2D strain was able to track the high velocities in the basal LV. A recent stress echo study did not find any difference in the diagnostic accuracy of 2D strain compared with a TDI-based method. 17 Interestingly, the agreement between ST-7P and TDI&ST was better than between TDI&ST and TDI-VG (Figure 3). Although ST-7P and TDI&ST were both based on segmentlength measurements, and thus very related, the similar results suggest that there is potential for further development of methods combining information from the two modalities to improve accuracy. The differences in agreement at the apical, mid-ventricular and basal levels suggest that different methods have advantages at different levels, which is important in further development. Previous studies have shown better or similar agreement between long-axis strain measured by echocardiography and MRI tagging. 7,11,18 Better agreement was probably due to a more comprehensive approach in acquisition and analysis of the MRI images, including interpolation in 3D. 18 In the present study we did not include or exclude patients based on image quality. Systolic and early diastolic strain rate Due to the derivation from strain, SR is inherently noisier than strain. Therefore it was expected that the correlations between MRI tagging and the echo methods were poorer for SR than for strain. 2D strain seemed to correlate better with MRI tagging than the other echo methods, and 2D strain and TDI-VG were the only methods to demonstrate a base-to-apex gradient in E SR of similar magnitude as in MRI (Table 2C). This gradient has not been found previously. 19 In MRI tagging, the low basal values were possibly due to poor tracking during early diastole. In 2D strain the 2D nature of the ROI causes the centre line of the ROI to shift inwards and outwards due to apical wall thickening and thinning, and this could have caused the higher apical values compared with the other echo methods. In TDI-VG the high apical values could have been caused by near-field reverberations. However, the fact that the difference between the mid-ventricular and apical level was not evident for S SR suggests that it was not due to methodological differences. The requirement of high temporal resolution for accurate measurements of E SR was illustrated by the fact that TDI-VG, which had the highest frame rate and least smoothing, measured higher values than the other methods. 20 MRI tagging was performed with 40 frames/s, which seemed adequate for strain and S SR. TDI&ST measured lower values of E SR than TDI-VG due to averaging of the tissue Doppler data between each B-mode frame. The correlation coefficients between echo and MRI tagging were lower for SR than for strain. However, the acceptable results for the comparisons of SR between tissue Doppler and ST methods suggest that ST methods, especially 2D strain, can give clinically useful SR measurements if the B-mode frame rate is high enough (Figure 4 and 5). Reproducibility The moderate reproducibility was probably an important cause of the variation between the methods. The better reproducibility in 2D strain was probably due to the large number of kernels along the whole LV wall, the automated positioning and segmentation of the ROI, and the applied smoothing. The better reproducibility of 2D strain was not reflected in better agreement with MRI tagging. We found that the reproducibility of MRI tagging was in the same range as for the echo methods, except 2D strain. In addition, the SD in E SR from tagging was only slightly higher than of S SR. This was probably caused by the longer T1-relaxation time on a 3T compared with a 1.5T system, which lengthens the duration of the tagging pattern. However, susceptibility-artefacts and eddy currents are prominent on a 3T system, and reduce the advantage of the improved signal-to-noise ratio. Quality assessment of speckle tracking results The tracking-quality test in 2D strain had reasonable accuracy, and could give the user important feedback if it

8 236 B.H. Amundsen et al. could be performed during scanning. The two markers of tracking quality extracted from ST-7P seemed to have potential, and cut-offs and possible combinations of the two should be tested prospectively. Sources of error in speckle tracking Reverberations impair tracking of the underlying myocardial speckle pattern (see Supplementary material online, Video S1). Dual-focus acquisition reduces near-field reverberations but reduces frame rate, which increases the risk of speckle decorrelation and poor tracking. Lateral resolution is intrinsically lower than resolution in the beam direction in a B-mode image. Lateral resolution is also depth-dependent, and is reduced if ribs or lungs reduce the effective probe aperture, or if frame rate is set to high (see Supplementary material online, Video S1 S3, Figure S1). Poor lateral resolution will make ST angle dependent, but as long as the segment long-axis direction is defined by the ROI, ST long-axis measurements will not be spoiled by the simultaneous wall thickening in the same way as tissue Doppler VG methods. Limitations Manual analysis could have improved the feasibility of TDI-VG, but our group has previously shown that this requires more time and experience than automated methods. 6 This previous study also showed that values obtained by manual analysis are closely related to those obtained by the automated methods. The MRI tagging method has not been validated for long-axis measurements, and hence cannot be said to represent a gold standard. However, MR tagging has been used previously to validate longitudinal strain measurements by echocardiography, 7,11,18 and measurements from a separate independent imaging modality was considered especially important in this study where many different echo methods were compared. The peak-combining harmonic phase software has been compared with conventional harmonic phase analysis for circumferential measurements, and found to reduce the variation in strain analysis. 13 Only one of the applied echo methods is commercially available (2D strain). However, we consider the results of the other three methods important because they illustrate the basic properties of different approaches to regional deformation analysis (ST/tissue Doppler). Measuring deformation by tracking segment end-points can give different results compared with measurements obtained along the curvature of the ventricular wall, which is more physiologically correct. Further, as strain in the myocardium is 3D, all the methods in the study have an important limitation, as they did only measure deformation in 1D. In the statistical analyses we did not adjust for the fact that segmental values from the same subject are not independent. Using a nested approach would have improved the accuracy in the analysis, especially for the ANOVAs. Conclusion ST alone or combined with TDI shows better agreement with MRI tagging for regional myocardial strain than measurements based solely on tissue Doppler VG. Such methods also seem to have potential for automated measurements of SR. Although the clinical significance of the detected differences between the methods needs to be established, the present study gives important information on the strengths and weaknesses of the different methods, which is important for further development to increase accuracy and clinical applicability. Supplementary data Supplementary data are available at European Journal of Echocardiography online. Funding B.H.A. and J.C. received funding from the Norwegian University of Science and Technology. The study was also funded by the Department of Medical Imaging, St. Olavs Hospital and Fund for Cardiac Research, Trondheim, Norway. Acknowledgement We appreciate the technical assistance of Jørn Kværness (Philips Medical Systems) with the MRI tagging measurements. Conflict of interest: B.H.A., A.S., and H.T. have received lecture fees from GE Vingmed Ultrasound, Horten, Norway. References 1. Heimdal A, Stoylen A, Torp H, Skjaerpe T. Real-time strain rate imaging of the left ventricle by ultrasound. J Am Soc Echocardiogr 1998;11: Weidemann F, Jamal F, Kowalski M, Kukulski T, D 0 hooge J, Bijnens B et al. Can strain rate and strain quantify changes in regional systolic function during dobutamine infusion, B-blockade, and atrial pacing implications for quantitative stress echocardiography. J Am Soc Echocardiogr 2002;15: Hanekom L, Jenkins C, Jeffries L, Case C, Mundy J, Hawley C et al. Incremental value of strain rate analysis as an adjunct to wall-motion scoring for assessment of myocardial viability by dobutamine echocardiography: a follow-up study after revascularization. Circulation 2005;112: Ingul CB, Rozis E, Slordahl SA, Marwick TH. Incremental value of strain rate imaging to wall motion analysis for prediction of outcome in patients undergoing dobutamine stress echocardiography. Circulation 2007;115: Marwick TH. Measurement of strain and strain rate by echocardiography: ready for prime time? J Am Coll Cardiol 2006;47: Ingul CB, Torp H, Aase SA, Berg S, Stoylen A, Slordahl SA. Automated analysis of strain rate and strain: feasibility and clinical implications. J Am Soc Echocardiogr 2005;18: Amundsen BH, Helle-Valle T, Edvardsen T, Torp H, Crosby J, Lyseggen E et al. Noninvasive myocardial strain measurement by speckle tracking echocardiography: validation against sonomicrometry and tagged magnetic resonance imaging. J Am Coll Cardiol 2006;47: Langeland S, D 0 hooge J, Wouters PF, Leather HA, Claus P, Bijnens B et al. Experimental validation of a new ultrasound method for the simultaneous assessment of radial and longitudinal myocardial deformation independent of insonation angle. Circulation 2005;112: Rappaport D, Adam D, Lysyansky P, Riesner S. Assessment of myocardial regional strain and strain rate by tissue tracking in B-mode echocardiograms. Ultrasound Med Biol 2006;32: Toyoda T, Baba H, Akasaka T, Akiyama M, Neishi Y, Tomita J et al. Assessment of regional myocardial strain by a novel automated tracking system from digital image files. J Am Soc Echocardiogr 2004;17: Cho GY, Chan J, Leano R, Strudwick M, Marwick TH. Comparison of twodimensional speckle and tissue velocity based strain and validation with harmonic phase magnetic resonance imaging. Am J Cardiol 2006;97:

9 Regional myocardial long-axis strain and SR measurements Aase SA, Stoylen A, Ingul CB, Frigstad S, Torp H. Automatic timing of aortic valve closure in apical tissue Doppler images. Ultrasound Med Biol 2006;32: Ryf S, Tsao J, Schwitter J, Stuessi A, Boesiger P. Peak-combination HARP: a method to correct for phase errors in HARP. J Magn Reson Imaging 2004; 20: Bland JM, Altman DG. Measuring agreement in method comparison studies. Stat Methods Med Res 1999;8: Ingul CB, Stoylen A, Slordahl SA, Wiseth R, Burgess M, Marwick TH. Automated analysis of myocardial deformation at dobutamine stress echocardiography: an angiographic validation. J Am Coll Cardiol 2007;49: Korinek J, Kjaergaard J, Sengupta PP, Yoshifuku S, McMahon EM, Cha SS et al. High spatial resolution speckle tracking improves accuracy of 2-dimensional strain measurements: an update on a new method in functional echocardiography. J Am Soc Echocardiogr 2007;20: Hanekom L, Cho GY, Leano R, Jeffriess L, Marwick TH. Comparison of two-dimensional speckle and tissue Doppler strain measurement during dobutamine stress echocardiography: an angiographic correlation. Eur Heart J 2007;28: Edvardsen T, Gerber BL, Garot J, Bluemke DA, Lima JA, Smiseth OA. Quantitative assessment of intrinsic regional myocardial deformation by Doppler strain rate echocardiography in humans: validation against three-dimensional tagged magnetic resonance imaging. Circulation 2002;106: Kowalski M, Kukulski T, Jamal F, D 0 hooge J, Weidemann F, Rademakers F et al. Can natural strain and strain rate quantify regional myocardial deformation? A study in healthy subjects. Ultrasound Med Biol 2001; 27: Slordahl SA, Bjaerum S, Amundsen BH, Stoylen A, Heimdal A, Rabben SI et al. High frame rate strain rate imaging of the interventricular septum in healthy subjects. Eur J Ultrasound 2001;14: