We evaluated the ability of a novel automatic index based on area strain to reliably quantify global and regional left ventricular (LV) function and accurately identify wall motion (WM) abnormalities using three-dimensional speckle tracking echocardiography.
A total of 140 consecutive patients underwent two- and three-dimensional echocardiography. Segmental WM assessment by area strain was compared with visual assessment of two-dimensional images by two experienced echocardiographers. For global LV function assessment, area strain was validated against LV ejection fraction (EF) and wall motion score index (WMSI). Observer reliability was assessed in all patients, whereas test–retest reliability was evaluated in a subgroup of 50 randomly selected patients. Normal reference values of area strain were determined in 56 healthy subjects.
Agreement of WM scores between area strain and visual assessment was found in 94% of normal, 55% of hypokinetic, and 91% of akinetic segments ( κ -coefficient 0.88). Sensitivity, specificity, and accuracy of area strain to distinguish abnormal segments from normal segments were 91%, 96%, and 94%, respectively. In regard to global LV function assessment, area strain was highly correlated with EF and WMSI ( r = 0.91 and 0.88, respectively). Observer and test–retest reliability of area strain for quantitative assessment of global and regional LV function were good to excellent (all intraclass correlation coefficients ≥0.77). Intraobserver and interobserver reliability of semiquantitative segmental WM analysis by area strain ( κ -coefficients 0.87 and 0.73) were comparable to visual assessment by experienced echocardiographers (0.85 and 0.69, respectively).
Area strain represents a promising novel automatic index that may provide an accurate and reproducible alternative to current echocardiographic standards for quantitative assessment of global and regional LV function. Area strain seems to adequately identify regional wall motion abnormalities compared with the clinical standard of visual assessment by experienced echocardiographers.
Accurate and reproducible evaluation of global and regional left ventricular (LV) function is of vital importance for the determination of diagnosis, prognosis, and therapeutic options of multiple cardiovascular diseases. However, in daily clinical practice the echocardiographic evaluation of regional LV function is mainly performed by visual estimation on two-dimensional echocardiographic images, which is known to be subjective and insufficiently reliable for sequential use.
The newly developed three-dimensional speckle tracking echocardiography (STE) provides a fast and comprehensive quantitative assessment of LV myocardial dynamics in all four dimensions, and does so with all LV segments in their spatial and temporal relation to each other within the same data set. With the development of three-dimensional STE technology, area strain was introduced as a novel automatic index for quantitative echocardiographic evaluation of global and regional LV function. During LV contraction, the endocardial surface area decreases in size because of longitudinal and circumferential shortening, and radial myocardial thickening. Area strain reflects this change in the endocardial surface area and quantifies it by giving the percentage change in area from its original dimensions ( Figure 1 ).
Because the change in endocardial surface area should be related to wall motion (WM) and the simultaneous change in LV volume used for the respective measurements of the wall motion score index (WMSI) and LV ejection fraction (EF), a strong relationship between area strain and these traditional parameters of global LV function is expected. More important, this parameter should be sensitive for detecting attenuating effects of ischemia and scar on regional WM, which are commonly most pronounced in the subendocardial layer of the myocardium. Thus, it may provide a fast and reproducible automatic assessment of wall motion abnormalities (WMAs) that is as accurate as visual assessment by experienced echocardiographers.
The primary objectives of this study were to (1) validate area strain with traditional parameters of global and regional LV function, in particular its ability to identify WMA compared with visual assessment by experienced echocardiographers; (2) determine its observer and test–retest reliability in patients with cardiac disease with a wide range of LV function; and (3) establish normal reference values in a healthy population.
Materials and Methods
Patients and Healthy Subjects
A total of 140 consecutive patients visiting our laboratory for echocardiographic examination were enrolled in this study. Twenty-six patients (19%) were excluded after echocardiographic acquisitions because of poor image quality (defined as >4 non-visualized segments) ( n = 16), irregular heart rhythm ( n = 7), or a failure in data exportation from the scanner ( n = 3). Of the remaining 114 patients in the study group (67 men, mean age 59 ± 16 years, mean LV EF 51% ± 13%), 48 had ischemic heart disease, 44 had various diagnoses of heart disease (e.g., valvular, congenital, and nonischemic forms of cardiomyopathy), and 22 had suspected cardiac disease, but no cardiac abnormalities were identified during echocardiographic examination.
To determine normal reference values for area strain, 56 selected healthy subjects were included in the study. The healthy subjects (44 men, mean age 40 ± 15 years, mean LV EF 61% ± 4%) satisfied the following criteria: no history of cardiac symptoms, hypertension, or diabetes; no use of medication; and normal physical examination, electrocardiogram, and echocardiogram results. All subjects gave informed consent, and the local ethics committee approved the protocol.
Three-dimensional STE imaging was performed from an apical position using a commercial scanner (Artida 4D, Toshiba Medical Systems, Tustin, CA) with a fully sampled matrix array transducer (PST-25SX). Wide-angled acquisitions were recorded, in which six wedge-shaped subvolumes were acquired over seven consecutive cardiac cycles during a single breath-hold. While retaining the entire LV within the pyramidal volume, depth and sector width were decreased as much as possible to improve the temporal and spatial resolution of the images, resulting in a mean temporal resolution of 21 ± 2 volumes per second.
A standard two- and three-dimensional echocardiographic examination was performed using a different commercial scanner (iE33, Philips, Amsterdam, The Netherlands). Two-dimensional echocardiographic images were optimized for segmental WM assessment by modifying the gain, compress, and time-gain compensation controls, after which cine-loops of three consecutive beats were recorded, while making an effort to avoid foreshortening. The methodology of acquiring three-dimensional echocardiographic data sets was similar to that of three-dimensional STE data sets with the exception of acquiring seven wedge-shaped subvolumes during a single breath-hold instead of six subvolumes. Three-dimensional STE images were analyzed online, whereas two- and three-dimensional echocardiographic images were stored digitally for offline analysis of segmental WM and LV EF, respectively.
Segmental Wall Motion Analysis
Segmental WM was visually graded in parasternal and apical two-dimensional views by two experienced readers (MFAA, OK) according to the appropriate 16-segment model and scored as follows: normal or hyperkinesis = 1, hypokinesis = 2, akinesis = 3, dyskinesis = 4, and aneurysmal = 5. WMSI was calculated as the average score of all analyzable segments. Because the number of dyskinetic and aneurysmatic segments was small, they were grouped together with akinetic segments for segmental comparisons of area strain versus visual assessment. Segments with agreement between both readers on WM scoring by visual assessment were used to determine the accuracy of area strain.
Segmental WM analysis by three-dimensional STE involved the readers (SAK, MFAA) to set three markers on two orthogonal apical views, namely, two markers at the edges of the mitral valve ring and one marker at the LV apex. The LV endocardial border was then automatically detected by the three-dimensional WM tracking software (Toshiba Medical Systems), after which the reader could manually adjust the endocardial border and myocardial thickness if necessary. The system then automatically performed the segmental WM analysis through the entire cardiac cycle, providing continuous values of global and segmental strain and displacement including area strain for all 16 segments simultaneously. Readers were blinded to the clinical data of the patients and the results of the other analysis method used.
Comparison of WM Analysis Methods
First, segments were evaluated on interpretability based on visual inspection of two- and three-dimensional images. Subsequently, receiver operating characteristic curve analysis was applied to determine optimal cutoff values for area strain to best differentiate among normal, hypokinetic, and akinetic segments as established by visual assessment (i.e., normal vs. other segments and akinetic versus other segments respectively). Analysis of agreement between area strain and visual assessment for identifying segmental WMA was done for three-level ordinal scores (normal, hypokinetic, and akinetic) and binary categorization into normal and abnormal segments. Sensitivity, specificity, positive and negative predictive values, and accuracy of area strain versus visual assessment were determined for all segments that were scored by both methods. This was done for the entire LV, and for basal, mid, and apical levels separately.
Observer and Test–Retest Reliability
Observer reliability of global and regional area strain was assessed in all 114 patients. Three-dimensional STE datasets were analyzed for interobserver reliability by two separate observers. Intraobserver measures were performed on average 1 week apart in random order. In addition, test–retest reliability was assessed in a randomly selected subgroup of 50 patients. A complete restudy was performed within 1 hour after the first study without alteration of hemodynamics or therapy. Each study was subsequently analyzed by separate observers to best reflect daily clinical practice.
Data were analyzed using SPSS version 15.0 (SPSS, Inc., Chicago, IL). Continuous data are presented as mean ± SD. Categoric data are presented as a count and percentage. Variables were analyzed for a normal distribution based on descriptives of mean, median, SD, minimum and maximum, and graphic inspection of histograms. Statistical significance was defined as P < .05.
The association between area strain and LV EF and WMSI were tested using intraclass correlation coefficients (ICCs). Differences in the presence of WMA in patients according to area strain and visual assessment were assessed with the Fisher exact test. Significance of differences in magnitude of segmental area strain values among normal, hypokinetic, and akinetic segments as determined by visual assessment was assessed by independent-samples t tests. To adjust for clustering of segments within subjects, multilevel analysis was performed using MlWin version 2.21. Intermethod agreement of WM scores between area strain and visual assessment was evaluated using Cohen’s unweighted κ coefficient for binary variables (normal vs. other, hypokinetic vs. other, akinetic vs. other) and the κ coefficient with quadratic weighting for the three-level ordinal variable (normal = 1, hypokinetic = 2, akinetic = 3).
Observer and test–retest reliability of area strain as a continuous variable were assessed using ICC with a variance components procedure (restricted maximum likelihood method of estimation), where the observer and subject (or segment) were entered as random effects. The clinical significance of ICC was interpreted as follows: excellent, ICC ≥0.80; good, 0.60 ≤ ICC <0.80; moderate, 0.40 ≤ ICC<0.60; and poor, ICC <0.40. The kappa coefficient with quadratic weighting ( κ qw ) was used as an index of interobserver and intraobserver reliability of area strain as a three-level ordinal variable (normal = 1, hypokinetic = 2, akinetic or worse = 3), and its clinical significance was interpreted in a similar manner as described for ICC.
Normal Reference Values
In the population of healthy subjects, global and segmental area strain demonstrated normal distributions and small SDs, indicating relatively tight normal ranges ( Figure 2 ). However, functional non-uniformity in the average value of area strain was observed between individual segments and between different regions and levels of the LV. In the circumference, area strain increased significantly from inferoseptal to anterolateral regions ( P < .05). Longitudinal non-uniformity was most pronounced on the apical level, demonstrated by higher values of area strain found in the apex compared with basal and mid-ventricular levels ( P < .001).
Regional LV Function Assessment
A total of 1824 segments from 114 patients were analyzed for regional WMA by means of visual assessment of two-dimensional echocardiographic images and automatic area strain assessment of three-dimensional STE volumes. Because of poor image quality, significantly more segments were deemed uninterpretable by automatic assessment of three-dimensional volumes than by visual assessment of two-dimensional images (223 [12.2%] vs. 63 [3.5%] segments, P < .001). Segments deemed uninterpretable in two-dimensional echocardiographic images were largely also considered uninterpretable in three-dimensional views, and the distribution of these segments was evenly distributed among ventricular levels. For three-dimensional echocardiographic in general, uninterpretable segments were primarily located in the anterior wall and apex (92% of all uninterpretable segments). After exclusion of uninterpretable segments and segments with equivocal scores by visual assessment (479 [26.3%] segments), 1345 segments (73.7%) were available for comparison of both assessment methods ( Table 1 ).
Area strain was significantly decreased in segments with WMA compared with normal segments, as well as in akinetic segments compared with hypokinetic segments ( Figure 3 ). To differentiate among normokinetic, hypokinetic, and akinetic segments, receiver operating characteristic curve analysis revealed optimal cutoff values of −32% and −24%, respectively. By applying these cutoff values, general concordance of WM scores by area strain and visual assessment was excellent ( κ qw 0.88), with agreement found in 94% of normokinetic segments and 91% of akinetic segments, but in only 55% of hypokinetic segments.
To distinguish abnormal segments from normal segments, area strain had sensitivity, specificity, positive and negative predictive values, and accuracy of 91%, 96%, 92%, 95%, and 94%, respectively. Equal performance measures for differentiation of normal, hypokinetic, and akinetic segments, each versus all other segments, are provided in Table 2 for the entire LV, and for basal, mid, and apical levels individually. Overall, area strain showed good performance versus WM scoring by visual assessment with a trend toward improvement from basal to apical levels (base: κ qw 0.84; mid: κ qw 0.89; apex: κ qw 0.92). However, the sensitivity and positive predictive value for detection of hypokinetic versus other segments were modest.
|Normokinetic vs. other||94%||94%||97%||86%||94%||0.855 (0.822–0.887)||0.977|
|Hypokinetic vs. other||55%||94%||62%||92%||87%||0.510 (0.440–0.580)||0.825|
|Akinetic vs. other||91%||94%||69%||99%||93%||0.747 (0.693–0.800)||0.975|
|Normokinetic vs. other||93%||90%||96%||82%||92%||0.802 (0.740–0.865)||0.967|
|Hypokinetic vs. other||53%||93%||59%||91%||86%||0.475 (0.364–0.587)||0.857|
|Akinetic vs. other||88%||94%||61%||99%||94%||0.683 (0.576–0.790)||0.975|
|Normokinetic vs. other||94%||96%||98%||87%||94%||0.873 (0.824–0.922)||0.982|
|Hypokinetic vs. other||55%||94%||63%||91%||87%||0.513 (0.402–0.624)||0.835|
|Akinetic vs. other||93%||93%||66%||99%||93%||0.732 (0.645–0.819)||0.971|
|Normokinetic vs. other||95%||96%||98%||91%||95%||0.899 (0.844–0.955)||0.984|
|Hypokinetic vs. other||61%||95%||65%||94%||90%||0.571 (0.424–0.719)||0.759|
|Akinetic vs. other||91%||95%||82%||98%||94%||0.821 (0.736–0.905)||0.979|