In the current work, we illustrate how two-dimensional deformation data obtained from echocardiography can be reconstructed into a 3-D model of the left ventricle in order to enable a more comprehensive description of mechanical activation and deformation dyssynchrony. Such an approach enables mapping of the spatio-temporal distribution characteristics of dyssynchrony and allows the implementation of newer indices aimed at estimating the impact of dyssynchrony on global ventricular performance. Using a model of mild dyssynchrony (WPW), severe dyssynchrony (LBBB) and an intervention on the dyssynchronous substrate (CRT), differences between and potential advantages of certain approaches are further discussed.
Differences between approaches to express dyssynchrony and dyscoordination
The most widely used method to describe dyssynchrony consists of measuring differences in timing of onsets and/or peaks of deformation throughout the ventricle [
10,
22,
23]. However, multiple shortening waves are very common in the dyssynchronous ventricle, making this method vulnerable to noise and rendering uniform definitions on "onsets" and "peaks" more cumbersome [
10,
22,
23]. When spatial information is encoded in the data-set, delays within the ventricle can also be expressed in terms of their spatio-temporal distribution patterns, e.g. by vector-analysis [
10,
19]. This approach may be preferable as it offers additional data on the organizational pattern of deformation and makes the analysis less vulnerable to accidental outliers or random noise in the measurements. In the present study, the additional value of vector analysis became apparent in the WPW-ventricle. A distinctly different spatio-temporal pattern of mechanical activation compared to the normal ventricle could be demonstrated, while in this small group no differences were detectable in variability of mechanical activation. Moreover, in the patients with left bundle branch block, CRT had a stronger effect on the vector magnitude than on the standard deviation of peak shortening timing [
10].
Myocardial dyssynchrony does not only induce heterogeneity of deformation-timing but also of -amplitudes. Because global ventricular function relates to global deformation, [
24,
25] and global deformation in turn depends on the deformation magnitude in the individual wall segments as well as on the coordination (synergy) between them, timing alone does not necessarily reflect the impact of the disturbance. Nelson et al. used CVeS as a marker of dyssynchrony in patients with idiopathic dilated cardiomyopathy and demonstrated that CVeS strongly predicted the acute benefit of CRT [
13]. However, not only dyssynchrony but also regional ischemia or scarring can affect end-systolic strain variance [
26]. Accordingly, in a recent study involving ischemic and non-ischemic patients, this parameter seemed less valuable [
27].
VSR represents a novel way to estimate the impact of dyssynchrony on global function. By "weighing" the observed end-systolic strain to the peak strain it may somewhat compensate for the aforementioned shortcoming of CVeS. Indeed, the VSR-value is insensitive to timely peaking but hypokinetic contractile behaviour. Nevertheless, in the present study neither CVeS nor VSR significantly changed upon resynchronization (see next paragraph: rationale for ESR).
ISF is a another new approach to estimate the impact of dyssynchrony by reflecting the part of the total deformation that is lost internally due to
simultaneous shortening and stretching because of dyssynchrony [
20]. ISF is rather a dyssynergy (= dyscoordination) than a dyssynchrony marker since it regards "synchrony of contraction" as "simultaneous shorten
ing or lengthen
ing in all parts of the ventricle". When different wall segments are deforming in phase with each other, there's synergy and ISF will be zero regardless of differences in velocity and extent of deformation. However, when some wall segments are deforming out of phase, the velocity and extent of their abnormal deformation do determine ISF. Hence, ISF becomes independent of the choice of peaks while remaining sensitive to strain amplitude differences of dyssynchronous segments. Preliminary results with ISF of circumferential shortening obtained by MR-T suggest this index of segmental interaction to be better related with long term remodelling than timing parameters only [
20]. Of interest, the present study indicates that CRT improves global ventricular function (GejS, ejection fraction) by a reduction of ISF, i.e. by a conversion of internal into external shortening. The fact that spatial distribution patterns cannot be deducted from ISF and that its value can be affected by random noise may represent limitations; with a small adaptation of the algorithm however, 3-D vectors of out-of-phase or paradoxical strain behaviour can be calculated throughout the cardiac cycle (See Additional file
2: Algorithm for ISF and vector of paradoxical strain-rate behavior (PSrV) and Additional file
3: Additional figure showing PSrV plot in NL and WPW). This fell beyond the scope of the present work.
The advantages of the proposed method in comparison with other techniques are summarized in the table attached in the appendix of this document (Additional file
4: Table 4: Comparison of commonly used echocardiographic techniques/indices to evaluate mechanical dyssynchrony with STOUT-indices)
Dyssynchrony analysis by myocardial deformation: unmet challenges
A key issue in the treatment of mechanical dyssynchrony is that electrical therapies-like CRT – can only amend electrical dyssynchrony [
28]. Unfortunately, heterogeneity of deformation and mechanical dyssynchrony are not always caused by electrical dyssynchrony [
29]. An imbalance in active and passive forces, causing deformation heterogeneity, can also occur in the absence of electrical activation delays [
26]. One of the true challenges for deformation imaging therefore lies in the distinction between mechanical dyssynchrony based on electrical dyssynchrony or based on other local conditions [
26,
30,
31]. In (local) pathologies such as ischemia for example, delayed and post-systolic shortening is rather a passive phenomenon of recoil than an expression of amendable dyssynchrony and premature shortening may represent a more specific marker. In addition, the relative amplitude changes in premature and delayed segments seen in animal experiments of acutely induced left bundle branch block suggest that postsystolic shortening in general may not represent shortening that can be recruited towards the end of the ejection period (see figure
5). Excluding post-systolic shortening from the analysis by confining measurements to the ejection period (e.g. by ISF) or by disregarding postsystolic peaks as in the calculation of ESR may therefore improve the estimation of truly recoverable dyssynchrony. In accordance with the latter hypothesis, VSR was not significantly changed by CRT in the present study, while ESR was significantly reduced.
Myocardial deformation or strain can reliably be measured in vivo by magnetic resonance tagging (MR-T) imaging [
24,
32]. This technique has been applied in animal models also during CRT, instigating the development of highly effective therapies like cardiac resynchronization therapy (CRT) [
3,
7,
10,
14,
15]. However, in humans MR-T has practical constraints and not all human's pathology can accurately be represented in animal models. Strain echocardiography by speckle tracking is a valuable alternative [
16,
33]. However, particularly in spherically dilated, thin walled and hypokinetic ventricles, the temporal and spatial resolution of echocardiography and the signal to noise ratio of speckle tracking are challenged. Frame rate, focus position and sector width can be adapted to optimize the ultrasound beam density and image quality in order to improve the reliability of speckle tracking [
16]. We therefore designed the current software in such a way that segmental data from single wall recordings can be imported separately if needed.
It has been recognized previously that the assumptions and algorithms used for data interpolation and incorporation into a 3-D model can alleviate but also introduce sources of error [
18]. However, all current deformation imaging modalities depend on reconstruction techniques and all are particularly vulnerable to grossly irregular heart rates. Because exact spatial location, orientation and geometry are known when MR-T is used, true 3-dimensional MR-T data sets can be reconstructed. Nor with the current, nor with a previously proposed echocardiographic methodology this is possible [
18]. With 3-D based speckle tracking software soon becoming available, the latter problem might be solved in the near future. Nevertheless, and in spite of using longitudinal instead of circumferential deformation, our ISF, dispersion, and vector data on mechanical activation and dyssynchrony closely resemble the published MR-T data in normal individuals and in patients with LBBB [
13,
19,
20,
34].
Echocardiographic strain analysis can be applied also in humans with contraindication to MR-T, such as following CRT. This has offered unique data on the effects of CRT in humans in the current and in previous studies [
35]. Finally, 2-DSE can measure deformation throughout the entire cardiac cycle and is thus independent of QRS triggering or fading of the taglines in diastole. This offers new opportunities. In the current work this is illustrated by providing the first preliminary data on mechanical activation vectors and dyssynchrony in WPW-patients.
Limitations
The presented echocardiographic approach remains time consuming and laborious, in particular related to the care taken to obtain high quality single wall recordings, the need for a meticulous registration of the timing events and the subsequent off-line calculation of deformation by the EchoPac-software at each of the wall segments individually. Once all files are transferred to STOUT however, little extra time is spent at the actual analysis of the traces and at making the dyssynchrony results available for statistical analysis. Another drawback of the methodology is that many of the presented indices will offer valuable information only when image quality is sufficient to provide robust deformation results covering most of the ventricle. In clinical practice, this can be problematic even when attempting to optimize quality by a single wall approach. The presented image acquisition and post-processing approach might therefore better serve research purposes than clinical practice but we expect newly gained insight to generate simpler methods for routine practice. One of such clinically more feasible methods to predict response to CRT for example, might be the calculation of ESR deducted from the septum only, as previous and the present work indicates that in LBBB the septal segments generally are the earliest (vector of peak time), display most stretching towards end-systole [
3,
7,
10,
14,
15] and thereby likely contribute most to the ESR-value.
Myocardial deformation is a complex three-dimensional event and differences in synchrony and synergy between the main axes of deformation have been suggested [
36]. In the present study we only reported on longitudinal deformation parameters. Transverse data from the same long-axis images and at the same location can be processed in STOUT, as well as circumferential, radial and transverse data. Although this allows a direct comparison, such study fell beyond the scope of the present work.
In the present study we primarily intended to highlight the differences (in strength) between the individual dyssynchrony indices and to point out some physiological aspects that have to be taken into consideration when expressing dyssynchrony and dyscoordination. Only a limited number of patients were therefore included in the present study. It is important to recognize that in recent literature many new technologies and dyssynchrony indices have been put forward, in some cases without providing either the pathophysiological rationale for their use nor a standardized methodology. In particular in the field of cardiac resynchronization therapy, many of them have entered the clinical arena long before being properly evaluated in multi-centre trials against simpler and user-friendly methods. Each new method should therefore be scrutinized regarding its rationale and tested for its feasibility and reliability in the real world. This is no different for the currently proposed indices; whether the higher sensitivity of a vector-, ISF- and/or ESR-based approach found in this study translates into a superior clinical yield remains to be established in larger, prospective studies.