Longitudinal peak strain detects a smaller risk area than visual assessment of wall motion in acute myocardial infarction

Twenty-six patients (23 men) average age 65 ± 8 years (range 50 - 78), table 1, were selected for this analysis from among 99 patients included in a study of primary PCI for ST-elevation myocardial infarction, from February 2006 to August 2007. The patients agreed to have acute echocardiography performed immediately as they were prepared for acute coronary intervention and on their return 4-8 weeks later for the determination of infarct size with MRI. Exclusion criteria were a new myocardial infarction (MI) and the need for rapid coronary artery bypass graft (CABG) surgery or renewed PCI during the interim period between primary discharge and the MRI. In the main study 159 patients were included and 99 finally completed the investigations. Acute echocardiography was possible only during office hours, enabling the inclusion of 26 patients for the echo substudy. All patients underwent coronary angiography with balloon dilatation and most frequently stenting, via a standard femoral approach. The culprit lesion was located in the left anterior descending coronary artery system (LAD) in 15 patients, in the right coronary artery (RCA) in 9 and in the left circumflex artery (CX) in 2 patients, table 1. Additional stenoses not dilated at the index event were seen in 10 of the 26 patients.

Approval was obtained from the Regional ethical review board in Linköping. The study complied with the Declaration of Helsinki and with agreements on Good Clinical Practice. All patients gave written informed consent.

At least three apical views were obtained with tissue Doppler (two-beat loops) while the patient was draped and prepared for acute PCI (GE V7 or GE V5 with 3 MHz transducers using harmonic imaging technology). At follow-up, a minimum of three apical loops were recorded, preferably in end expiration to minimize translational movement of the heart. The left ventricle was divided into 16 segments (6 basal, 6 mid, and 4 apical) [30]. The myocardial motion of each segment was evaluated according to the standard American Society of Echocardiography wall motion scoring system and WMSI was calculated [30, 31] (additional file 1). The three observers were blinded as to the temporal order of the echocardiographic investigations as well as to the result of the determination of transmurality.

Tissue Doppler images were acquired with a frame-rate exceeding 90/s. Off-line analysis was performed using Echopac software (Echopac BT08, GE Vingmed Ultrasound, Norway). In each of the three apical views, six segments were defined and regions of interest (ROI) using sample volumes of 6 × 12 mm were applied. Tissue Doppler values in the apical anteroseptal and inferoseptal segments were averaged into an apical septal segment and in the anterolateral and inferolateral averaged into an apical lateral segment thus allowing conversion from 18 segments into a 16 segment model. After checking for aliasing in the velocity mode, myocardial strain and displacement curves were drawn and the peak values, regardless of delay time, were measured on two consecutive beats, if possible (figure 1, additional file 2). Peak systolic strain was calculated from the velocity determination in the longitudinal direction. Normal shortening strain was denoted with negative values while lengthening was assigned positive values. Displacement from the base towards the apex was given negative values and displacement towards the base positive values [32] (additional file 3). Long axis left ventricular function was assessed from mitral annular motion (MAM) which was measured in four positions on the mitral annular ring. Values for strain, displacement and MAM were obtained by two observers and their average was calculated. The reference values for longitudinal strain by Kowalski et.al. (-16% ± 5%) were used as cut-off [20]. Left ventricular ejection fraction was calculated using the biplane Simpson's method of discs and averaged from two consecutive heart beats in exams with sufficient image quality.

To obtain global measurements of wall motion and strain, measurements from each segment were added and the sum divided by the total numbers of segments. To assess regional measurements of wall motion and strain in relation to infarct transmurality, numbers for the three segments of each wall (anterior, lateral, posterior and septal) were averaged.

The patients were placed in the magnet (1.5 T Achieva, Philips Healthcare, Best, The Netherlands) in supine position. A circular polarized body-array surface coil was used in all measurements. ECG-triggered MR images were obtained during repeated breath-holds.

Cine-MR imaging was performed with a balanced steady state free precession turbo field-echo (b-SSFP TFE) sequence that covered the entire left ventricle with on average 18 (range 10 - 23) short-axis slices and three long axis planes (apical 2-, 3- and 4-chamber views). Temporal resolution ranged between 26 - 41 ms (30 acquired phases). The inversion recovery turbo field echo (IR-TFE) sequence was a segmented 3D spoiled gradient echo sequence with TE = 1.3 ms, TR = 4.4 ms and TFE factor 43, leading to an acquisition phase time of 188 ms acquired during diastole. Slice thickness was 10 mm, intersection gap -5 mm (i.e. slices were overcontiguous), field-of-view 350 mm and image matrix 128 × 256. The contrast-enhanced images were acquired at the same slice positions as the cine-images, about 20 min after the administration of gadopentetate dimeglumine (Gd-DTPA) 0.2 mmol/kg bodyweight (Schering Nordiska AB, Järfälla, Sweden). Optimal contrast between hyperenhanced areas and normal myocardium was maintained by continually adjusting the inversion time to null the signal from the healthy myocardium.

Scar size was measured by two different observers on short-axis images using a freely available software "Segment" http://​segment.​heiberg.​se, [33]. Infarct volume and percentage was calculated from the short axis stack of slices (scar area × thickness × number of slices) and was averaged for the two observers. Infarct transmurality per segment was in this setting determined in the three apical views by using "Segment" and defined as segmental scar area (6 segments per view, figure 2). The 18 segments were reduced to 16 segments, identical to those used in the tissue Doppler analysis. Enddiastolic myocardial and cavity volumes were measured from the short axis LGE images (segmented myocardial or cavity area × thickness × number of slices). Myocardial mass (gram) was obtained by multiplying volume (ml) by 1.05

All heart-related parameters were reasonably well normally distributed and presented as mean ± (SD). The difference on global level between pre- and post PCI exams was analysed by two-sided t-test for paired observations. Spearman's rank correlation was used for global-, regional-, and segmental functional parameters vs. infarct size and infarct transmurality. The difference between normal segments (transmurality < 1%) and segments with transmurality ≥ 50% on regional level was analyzed with Mann-Whitney's U-test. Receiver-operator-characteristics (ROC) curve analyses were performed using the statistical software MedCalc^® Version 6.10 (MedCalc Software, Mariakerke, Belgium). The interaction between WMSI and strain on the detection of segments with a transmurality ≥50% was analysed with logistic regression.

Interobserver variability of the functional measures was expressed as standard error of a single determination (S_method) using the formula, first proposed by Dahlberg [34], S_method = √(∑d_i ²/(2n)), where d_i is the difference between the i:th paired measurement and n is the number of differences. S_method was also expressed as % over all means when applicable.

Left ventricular myocardial volume was 167.8 ± 36.9 ml (range 88 - 254 ml). Infarct size, determined at follow-up 4 - 8 weeks after PCI, was on average (± SD) 14.9 ± 5.6 ml (8.7 ± 7.4% of the volume of the left ventricular myocardium). Mean transmurality of affected segments was calculated for each of the 24 patients with follow-up scars ≥ 1%, giving a patient average of 44.3 ± 18.0% (range 8.5 - 76). At follow-up, WMSI improved from an average (± SD) of 1.6 ± 0.3 to 1.4 ± 0.3 (n = 26; p = 0.001).

Corresponding numbers for ejection fraction were 38.5 ± 8.5% to 46.8 ± 8.5% (n = 17; p = 0.001) and for global strain -13.2 ± 3.3% to -15.7 ± 3.5% (n = 26; p < 0.001). Displacement changed from -5.4 ± 1.8 mm to -6.1 ± 1.4 mm (n = 26; p = 0.030). The change in MAM did not reach statistical significance (mean 10.5 ± 2.7 mm to 11.1 ± 2.3 mm; n = 26; p = 0.167).

There were statistically significant correlations between infarct size and percent transmurality post-PCI (assessed by MRI) on the one hand and systolic ultrasonic measures pre and post PCI on the other (table 2). The highest correlation was for WMSI post PCI vs. infarct size (r = 0.83) and for WMSI post PCI and transmurality (r = 0.88), but also global strain post PCI correlated with infarct size (r = 0.51) and with transmurality (r = 0.64). Individual functional global measures pre- and post-PCI, ranked after scar size post-PCI, are shown in figure 3.

Wall motion score correlated moderately with transmurality in the acute phase and at follow-up (r = 0.67, p < 0.001 and r = 0.63, p < 0.001) while the correlation for longitudinal strain vs. transmurality was lower at both time points (r = 0.51, p < 0.001 and r = 0.44 p < 0.001) and for MAM still lower (r = -0.25, p = 0.01 and r = -0.41, p < 0.001). In a comparison between walls with transmurality ≥ 50% and normal walls, WMSI, strain and MAM all displayed values that were significant. The change post-pre was not significant for any of the three measures, see table 3.

In the initial study, 390 out of 416 segments (94%) were successfully visualised and at follow up 410 of the 416 segments (99%). Wall motion score correlated moderately with transmurality in the acute phase and at follow-up (r = 0.58, p < 0.01 and r = 0.53, p < 0.01) while longitudinal strain correlated weakly with transmurality at both time points (r = 0.38, p < 0.01 and r = 0.31 p < 0.01). Displacement, regardless of the position of the segment (apical-middle-basal) also correlated weakly with transmurality (r = 0.28, p < 0.001 and r = 0.34, p < 0.001).

The segmental values were analysed from two approaches. One was based on the evaluation of WMSI (normal or abnormal) and transmurality in regard to strain for segments that either deteriorated, improved or remained unchanged between the acute event and follow-up (table 4). In the other approach, normal or reduced strain was the watershed from which the level of transmurality and changes in WMSI were evaluated (table 5).

Initially, 231 out of 390 visible segments displayed wall motion abnormalities. 97 of these developed scar with transmurality > 1%, i.e. final scar affected 42% of the segments at risk (table 4). Based on longitudinal strain > -11%, 129 segments were abnormal initially of which 60 developed scar with transmurality > 1%, i.e. final scar affected 46% of the segments at risk (table 5). However, of the 261 segments that were determined to be normal according to strain < -11%, 47 segments (18%) developed scar compared to 10 segments (6%) with WMSI. The two methods were compared in a ROC-analysis as for the prediction of segments that were to develop scar with a transmurality ≥ 50%, figure 4. AUC was significantly higher for WMSI (0.92) than for strain (0.78), p < 0.0001. Sensitivity at 80% specificity was for strain 64% and for WMSI close to 90%. In a logistic regression analysis incorporating WMSI and strain, both parameters were significant for the prediction of transmurality ≥ 50% but strain did not add significant information beyond that carried by WMSI.

The calculated error (S_method) between three observers were for WMSI 0.50 (coefficient of variation (COV) = 32%) and 0.42 (COV = 30%) calculated for the pre- and post-PCI investigations. Corresponding methodological errors for the two observers of the MAM were 1.30 mm (12.4%) and 1.04 (9.3%), respectively. The methodological error in absolute numbers was for strain pre-PCI 5.4 and for post-PCI 4.5. Corresponding values for displacement were 1.59 and 1.43, respectively. COV for these measurements is of no interest since the values include zero. No significant differences were found, in regard to these measurements, between pre- and post PCI examinations. S_method for scar assessed by MR post-PCI was 1.6 ml (11%) or related to myocardial volume 1.1% (12%).

In this study we show that the initial evaluation based on visual assessment of wall motion as well as on objective measurements of wall deformation correlate with the final infarct size and transmurality at follow-up after 4 - 8 weeks. Functional evaluation based on echocardiography on the catheterization table seems valid, since those measurements correspond fairly well with assessments performed under less hurried conditions at follow-up. Prediction of segments/walls that will develop permanent scar is possible with both methods but visual assessment seems superior in this respect (figure 4). Visual wall motion assessment has been criticized because of its subjectivity [35] but was robust in the present study. Peak strain from myocardial tissue Doppler showed less correlation with final infarct size and transmurality than visual assessment, both in the acute study and in the quiet imaging conditions at follow-up.

Global strain reflects the averaged segmental myocardial long-axis relative shortening and is a global functional measurement that may give information beyond what is available from WMSI and left ventricular ejection fraction (LVEF) [36]. Strain is an objective measurement compared with the subjective evaluation of wall motion and might therefore be a valuable instrument in the daily workflow. However, we found that global strain showed a moderate correlation with total infarct size and mean transmurality, lower than that for WSMI. To the contrary, Gjesdal et.al. showed a higher correlation (0.84) between global strain and scar compared to WMSI and scar (0.70) in patients with chronic myocardial infarction [37]. In their study, 2D-speckle tracking echocardiography was used on somewhat younger patients (mean age 55 yrs) 9 months after MI. Additionally, Vartdal et. al found a correlation of 0.77 between strain and infarct size in patients with acute STEMI [38]. The corresponding figure for WMSI and infarct size was r = 0.45. However, these patients were examined 1.5 h after revascularization when ischemic wall motion abnormalities could have declined, possibly faster than changes in strain. Interestingly, they also found that global strain might be a valuable predictor for the total amount of scar and hence might be a clinical tool for risk stratification. Recently, Gjesdal. et.al. showed a significant correlation between MAM, measured in 4 positions, and infarct size (r = 0.58, p < 0.01) in patients with chronic scar 9 months post AMI [39]. Additionally, Sjøli et.al. showed a correlation of 0.62 between global strain and infarct size measured within 3.5 h after revascularization [40]. Global strain showed a higher correlation with the size of myocardial scar compared with LVEF. This is in line with our results showing that LVEF has only a weak correlation with infarct size and mean transmurality (table 2). Ugander et.al. also confirmed a large variation in LVEF in relation to infarct size and concluded that infarct size cannot be used to predict LVEF [41]. The number of dysfunctional adjacent segments seemed to be a more important determinant on regional wall function than infarct transmurality [42]. This could be caused by extensive hibernation or a compensatory increase in wall motion in healthy parts of scarred ventricles.

By averaging measurements from the three segment levels in each wall, regional measurements could be correlated with regional scar. This was done to allow a comparison of strain and wall motion scoring with the regional motion amplitude of the mitral annulus, which could easily be obtained from tissue Doppler velocities integrated over time ("tissue tracking"). We found a rather low correlation for regional MAM vs. transmurality, higher for regional strain and the highest for WMSI vs. transmurality.

Evaluation of wall motion is very important in the clinical setting but draw-backs are the subjectivity [35] and the extended learning curve for the physician. Strain has been proposed to identify acute ischemia [24] and to grade myocardial dysfunction [21, 32]. In this study we hypothesized that strain could quantify left ventricular function in relationship to scar transmurality. However, we found a higher correlation between wall motion and scar transmurality compared with strain vs. transmurality, as well as displacement vs. transmurality, both pre- and post PCI. Despite the theoretical advantages of strain, visual assessment by three experienced observers performed better when predicting scar transmurality in both the acute and chronic settings. Displacement, correlated even more weakly than strain (r = 0.28 pre, r = 0.34 post). This is in line the finding of Skulstad et.al. who studied the relationship of strain and displacement for quantification of regional myocardial function [32].

To determine myocardial salvage, an initial measure of myocardial area-at-risk is required. Myocardial oedema determined with MRI and myocardial perfusion determined with SPECT have been used, mainly because imaging can be delayed for several days for MRI and for 6-8 hours in regard to SPECT. In the present study, we selected wall motion assessment, either visual or by tissue Doppler, because of its low cost, availability at bedside, and the possibility to perform assessment in immediate relation to the intervention. We found that both methods identified segments that were to develop scar transmurality in excess of 50%. However, wall motion scoring performed better than strain measurement both in terms of detecting initially threatened segments and in avoiding false positive results. In total, 42% of threatened segments according to wall motion scoring developed significant myocardial scar. This result is in line with guidelines suggesting that the goal of acute infarct treatment is a final scar size < 40% of the initial risk area [12].

The first echocardiographic examination was performed under difficult scanning conditions, while preparing the patient for acute PCI resulting in a compromised image quality. However, correlations were in the same range as for measurements performed at follow-up, when the imaging conditions were favourable. Strain measurements are known to be load dependent [19, 43], which can have influenced measurements performed under the acutely unstable hemodynamic conditions of myocardial infarction. While velocity, displacement and strain obtained with tissue Doppler are angle-dependent, 2D-speckle tracking adds a second dimension to the scanning plane which reduces the influence of off-angle effects. Accordingly, speckle tracking MRI has shown higher values for peak longitudinal strain of healthy segments (about -20%) vs our tissue Doppler values (-17) [44]. In the present study, speckle tracking was not performed because the underlying gray scale frame rate in the colour tissue Doppler loops was too low to allow good quality tracking Tissue Doppler produces a strong signal from the myocardial wall also in difficult imaging conditions and was thus projected to produce a more stable result than visual assessment. But, in the post processing analysis, strain curves turned out to be sensitive to the placement of the ROI's as well as to shadowing from the lungs.

In this study we have used a 16 segment model of the left ventricle in favour of the newer 17 segment model proposed by AHA [45]. However, since the 16 segments are identical with those of the AHA model and measurements were not performed on the apical cap, the model was not considered to have an influence on the result.

Registration (= aligning images from different studies) is a major problem when studies are performed on patients at different time points and with different modalities. However, WMSI, strain, displacement and infarct transmurality were all assessed in the long axis view and care was taken to align the imaging plane with that used for measuring transmurality.