Background
Cardiovascular magnetic resonance (CMR) is the gold standard for the assessment of regional myocardial strain, which provides a sensitive and quantitative indicator of myocardial function and viability[
1,
2]. Several CMR techniques including but not limited to phase-contrast velocity-encoded[
3], tagging[
4,
5], displacement encoding with stimulated echoes[
6], and strain encoding[
7] imaging are developed to accurately measure one-dimensional (1D) or two-dimensional (2D) strain. These methods have been validated in phantoms, animals, healthy subjects, and patients, and have demonstrated high accuracy and reproducibility in assessment of regional strain[
8]. Since 1D and 2D assessment of regional strain requires simplified geometric assumptions on cardiac geometry and deformation, three-dimensional (3D) approaches have more recently been developed and reported to be in agreement with conventional 2D techniques. However, the first 3D approaches required prolonged examinations or sophisticated analysis methods often with extensive manual interactions[
8]. Although recent improvements in these techniques, e.g.[
9‐
12], provide better solutions, common practice is still focused on 1D strain (predominantly circumferential strain)[
13‐
16] or 2D strain (predominantly circumferential and radial strains)[
17‐
19], in part due to undocumented benefit of fully 3D strain analysis.
A primary purpose of the present study was to test the hypothesis that quantification of 3D strain is superior to that of 1D (circumferential) and 2D (circumferential and radial) strain quantifications in depiction of regional mechanics and discrimination of viable and nonviable myocardial regions. Accordingly, we studied the contractile mechanics of an
in vivo porcine model of myocardial infarction (MI). We employed z-encoding harmonic phase (zHARP)[
20] and late gadolinium enhancement (LGE) imaging to measure regional strain and myocardial viability, respectively. Univariate, bivariate, and multivariate analyses of regional strain were performed to investigate the diagnostic accuracy of 3D strain analysis compared with 1D and 2D analyses.
Discussion
Advances in CMR[
9‐
12] allow comprehensive 3D quantification of myocardial contractility without the simplified geometric assumptions commonly associated with 1D or 2D techniques. The clinical adoption of 3D methods however has been hindered in part due to increased imaging time and processing steps required by majority of such methods. Recent developments in CMR imaging and analysis aim to address these issues, and decrease the scanning and processing durations. zHARP, in particular, enables 3D tracking of short-axis slices using simultaneous in-plane and through-plane displacement encodings without affecting the duration of image acquisition. As a result, imaging multiple orientations is eliminated and zHARP yields 3D strain measures at every pixel in the imaged slices without the need for image registration or numerical interpolations[
27]. Using zHARP, the present study was designed to assess the value of 3D regional mechanics analysis over the usual in-plane analysis. This incremental value can potentially facilitate the clinical adoption of 3D methods.
In this study, we used a porcine model of MI[
34] in which transmural infarction was present in the apical to mid levels. Mean infarct volume was 16% of LV volume with no post-MI variations at 11 days and one month (p = NS). Additionally, classic features of post-MI remodeling was demonstrated in the wall and ventricular chambers (Table
1), accompanied with wall thinning in infarct segments, and wall thickening in remote segments (Figure
3). In the adjacent segments, which contained no transmural infarct and very little sub-endocardial infarct, the wall thickness measurements showed no thickening at early post-MI. Wall thickening in these segments however improved at the late time point. Unlike previous research that reports a decrease in ejection fraction after LV remodeling, an increase in ejection fraction was observed at late post-MI. This observation can be due to mitral valve regurgitation.
For regional strain assessment, the myocardium was divided into segments based on the transmurality and distribution of infarct zone over the circumference of myocardium, characterized by LGE images (Figure
1). Unlike the standard 16-segment model, this method of segmentation eliminated segments with partial transmural infarcts and enabled us to independently study strain patterns in transmural infarcts and the peri-infarct regions that demonstrate different post-MI mechanical behaviors[
35]. The transmural infarct consistently spanned the anterior and anterior septal segments, as the infarct was generated following the same procedure and occlusion location in all animals. The size of infarction thereby was largely consistent with a relatively small variation among segment sizes at each cardiac level.
While 3D strain measures in remote segments did not vary significantly from that of the pre-MI healthy segments, the transmural infarct segments and their adjacent segments demonstrated significantly smaller values of 3D strain compared to the remote and healthy regions. Subsequently, univariate, bivariate, and multivariate regression analyses demonstrated that all the directional strain indices (E
cc
, E
rr
, and E
ll
) are significant covariates in identification of adjacent and infarct segments from healthy counterparts. The multivariate model using E
cc
, E
rr
, and E
ll
resulted in significant improvement (p < 0.01) in c-statistic and diagnostic accuracy (c-statistic = 0.981, accuracy = 96%) compared with the 2D model using E
cc
and E
rr
(c-statistic = 0.941, accuracy = 81%), and the univariate model using E
cc
(c-statistic = 0.846, accuracy = 71%). This finding shows that 3D strain may therefore allow a more accurate and reliable mapping of regional contractility in infarct neighboring regions in comparison with commonly employed 1D or 2D strain measures. However, the 3D strain analysis (c-statistic = 0.996, accuracy = 98%) did not appear significantly superior to simplified single-plane quantification of strain (2D: c-statistic = 0.991, accuracy = 94%, 1D: c-statistic = 0.986, accuracy = 95%) in the presence of a transmural infarct.
The 3D principal strain indices (
E
1
,
E
2
, and
E
3
) corresponded to the three directions of maximal myocardial deformation.
E
1
, corresponding to maximum thickening, was always larger than radial thickening not only in the infarct segments but also in the remote, and adjacent segments (Table
2). Similarly, the maximum shortening values,
E
3
, were larger (in absolute value) than the corresponding circumferential strain values in all segments at all time points. These principal strain indices demonstrated a precise assessment of regional function and did not illustrate inferior diagnostic performance in comparison with the more commonly used directional strain indices (Figure
6). The decomposition of principal strain indices from 3D strain tensor, unlike directional strain indices, does not depend on the gross geometry of LV and the orientation of imaged slice. This universal aspect of principal strains and their insensitivity to image planning may result in more accurate and robust strain values in comparison to
E
cc
,
E
rr
, and
E
ll
. However, this speculation requires a more careful study to be confirmed. The direction of principal strain may also provide a regionally varying index of myocardial deformation. However, the extraction of angles and averaging those angles within each segment must be performed under special care especially in regions with infarct, which may lack a dominant direction of deformation. Definitive conclusion warrants further investigation and analysis, which is beyond the scope of the present work.
CMR-based 3D strain may therefore allow accurate and reliable mapping of regional contractility and assessment of viability. 3D analysis may also have implications beyond analysis of myocardial infarction and it may help in assessing patients with ischemic or non-ischemic cardiomyopathy[
36]. The present study may help lay the road for a larger follow up study, with a longer post-MI time period and a wider spectrum of MI size, to investigate if an early strain signature can predict post-MI remodeling and determine the salvageable myocardium.
Limitations
The MI model used in this study was approximately 16% of the LV volume and restricted to the apex and septum. Furthermore, the MI was mainly transmural and the extent of sub-endocardial infarct was small (<25% area of adjacent segments). Infarctions in clinical practice that are not revascularized in time usually present a wider spectrum of MI size and distribution with the potential for larger sub-endocardial infarcts. The latter condition would present the ideal condition for 3D regional function analysis. The segmentation method for the baseline images and the post-MI images was different, which prevented direct comparison. Also, we did not perform pathologic validation of our model. However, this porcine model of MI has been characterized using high resolution
ex vivo CMR[
37] and has been reported to exhibit all the morphologic and functional remodeling features of clinical infarctions[
34]. There is also extensive published literature on the spatial correlation of infarct and its neighboring regions with delayed enhancement. Similarly, we did not perform a dobutamine challenge to assess the viable segments. This concept also has been validated repeatedly with regards to delayed enhancement. Lastly, we did not assess mitral valve regurgitation, which could potentially explain the increase in ejection fraction at the late post-MI time point.
Competing interests
Jerry L. Prince is a co-founder of Diagnosoft Inc., a company that licensed the HARP technology. The terms of this arrangement are managed by the Johns Hopkins University in accordance with its conflict of interest policies. Other authors have declared no competing interests.
Authors’ contributions
SS: substantial contributions to design of study, analysis and interpretation of data, and drafted the manuscript. KZA: substantial contributions to design of study, acquisition and interpretation of data, and revision of manuscript. TS: substantial contributions to acquisition of data. HA: substantial contributions to interpretation of data. MRA: substantial contributions to design of study, interpretation of data, and revision of manuscript. TPA: substantial contributions to design of study, interpretation of data, and revision of manuscript. JLP: substantial contributions to design of study, interpretation of data, and revision of manuscript. All authors read and approved the final manuscript.