During the past decade, technological progresses in non-invasive cardiac imaging (such as echocardiography and cardiac magnetic resonance) allowed significant improvements in our understanding of the physiology of ventricular mechanical contraction. Thus, nowadays, the ejection fraction (EF), based on a volumetric approach to systolic function analysis, is no longer considered the unique way of expressing the state of left ventricle (LV) function. Nevertheless, in daily practice, EF remains the most used parameter because of its simplicity and its reliability as a prognostic factor (1, 2). The development of deformation imaging, or strain imaging (SI), has led to a better understanding of ventricular mechanics and helped in the understanding of heart failure, especially in cases of diastolic dysfunction with preserved EF (3–5). The SI technology is based on tracking the tissue during the cardiac cycle in transthoracic echocardiography (TTE) by optical flow techniques (Speckle Tracking) (6, 7). The SI was secondary developed in cardiac magnetic resonance (CMR). Initially, various techniques were proposed, including myocardial tagging and strain-encoding (SENC), before the recent advancement of Feature Tracking (FT) technology (8–10). The primary advantage of using SI with CMR is the precision in positioning the short-axis (SAX), allowing for a more accurate assessment of GCS and GRS. This method is less reliant on the operator compared to TTE, thus enhancing its reliability (8, 11). Nowadays, FT has become the most widely used CMR technique as it allows for strain evaluation without the need for additional sequences. It allows the assessment of several parameters including the global longitudinal strain (GLS) describing the base-apex shortening, the global circumferential strain (GCS) describing the reduction of the mean diameter of the left ventricle and the global radial strain (GRS) describing the thickening of the left ventricular wall. In practice, GCS and GRS are less utilized than GLS because more difficult to assess in TTE, mainly due to the challenge of obtaining reproducible short axis slices (SAX). SI may be still underexploited in clinical practice although it is proven early prognostic factors in many cardiomyopathies both in adults and in pediatric populations (12, 13). One explanation could be that until recently, these parameters were evaluated independently of each other, giving therefore only a partial view of a complex mechanism. Indeed, these parameters may reflect the contribution to contraction of different myocardial layers while they both participate in overall cardiac mechanic efficiency (14). Associating deformation parameters and EF in a single mathematical relationship could be a lead to follow as Stokke et al. (15) and Pedrizzetti et al. (3, 16), introduced a concept of objective interdependency between volumetric and mechanical parameters (Central illustration). This new area of non-invasive exploration of cardiac function has the potential to emerge and to give a relevant space for SI in clinical practice (14). However, these theoretical works combine parameters that can be obtained from different modalities (echocardiography, CMR) and from either long-axis slices (LAX) or short-axis slices (SAX); Recommendations in CMR is evaluating LVEF and GCS from classical SAX slices while evaluation of GLS, that requires LAX views, is less frequent. Recently, CMR studies introduced the possibility to evaluate LVEF and the GCS using LAX, as well, by reconstructing the three-dimensional (3D) LV geometry by a combination of multiple LAX recordings (16).

The systematic evaluation and comparison of different FT-CMR measurement methods for these parameters (GCS, GLS, GRS and LVEF) in clinical practice has not yet been performed. This work aims to fill this gap by testing these different procedures during the normal clinical activity in CMR with a heterogeneous population of patient followed for various cardiovascular disease. Specifically, to compare the methods for the evaluation of the GCS and LVEF obtained from multi-slice SAX and tri-plane LAX, and how the different evaluations fit in the theoretical relationship between EF and global strain.

In the present study, which spanned one year and involved 134 patients, we illustrate the feasibility and reproducibility of CMR-derived SI parameters across a diverse spectrum of left ventricular (LV) morphologies in a heterogenous population.

Based on a systematic comparison between CMR measurements of LV function obtained by SAX and LAX slices performed with FT, we showed that these SI parameters were highly correlated but were not quantitatively interchangeable. Our work, performed in the setting of a heterogeneous population, intends to guide clinical practice by comparing the different techniques of acquisition of GCS, GLS, GRS and EF so as to ensure a strongest accuracy when employing different methods. This approach seems us important as functional parameters are used in daily practice as prognostic outcomes and can have an impact on therapeutic decision (20).

First of all, concerning EF calculation, the gold standard in CMR is still the manual SAX method (21). Semi-automatized EF-LAX performed with FT technology seems to systematically underestimate the ESV and overestimate the EDV and in consequence the EF by 7% in relative terms. This is in line with previous reports evaluating volumes and EF with CMR or TTE using manual or semi-automatized LAX method (21–25).

It is obvious that employing a semi-automated LAX method could result in substantial time savings. However, like any automated software, it does not fully consider the intricacies of abnormal left ventricle geometries. Indeed, in our cohort we faced two “risky” situations concerning almost the third of our cohort: in cardiomyopathy, the systolic borders can be harder to delimit in LAX than in SAX plane especially in systole where the papillary muscles can merge with the myocardial walls. Moreover, in congenital heart disease the left ventricle outflow track often has a complex geometry that has to be included in the 3D volume, which is not easily feasible with automated FT measures (26). Hence, we assumed that expert-performed manual contouring was the most reliable method for assessing volumes in especially in hypertrophied or twisted ventricles (27). However, this level of difference could be also explained by an insufficient manual correction of the left ventricle boundaries but the excellent correlation of EF-LAX in inter and intra-observer measures seems to exclude this theory. Finally, it depends also on the quality of the acquisition planes itself, and the precision of their 3D position relative to the ventricular anatomy.

Concerning the method of acquisition of GCS, we confirmed that GCS-SAX is systematically lower than GCS-LAX (16). This difference follows from the fact that GCS in SAX is taken on transversal slices that remain fixed in space while the LV contracts. As exemplified in Figure 6, during contraction the base moves apically and slices that are located at the LV base at end-systole became more apical at end-diastole. Therefore, a more basal and wider portion of myocardium enters the fixed slice, thus introducing an apparent widening and reducing the GCS. Differently, when measuring GCS in LAX images, the transversal slices move together with the tissue, thus they can cover the entire LV from base to apex uniformly during the entire heartbeat and give a closer representation of the physical contraction of the LV chamber. This phenomenon was previously mentioned in a very similar proportion between the two GCS measured on the endocardium in a different dataset (16). Of course, these phenomena would theoretically be better tackled in 3D cine sequences, but accessibility and acquisition time are still important limitations for the widespread application of strain in these sequences (24, 28). Finally, the usage rules of GCS with CMR should be clarified as a recent publication of the American society of echocardiography did it for TTE (29). Nevertheless, the fact that normal GCS values are potentially depending on sex, age and race is a barrier to its general utilization in clinical practice (29). Notably, the GRS exhibited poor inter- and intra-observer correlation and lacked association with EF. This discrepancy is likely attributed to challenges in automatically defining the epicardial contour across all SAX slices.

Our findings reveal strong correlations between individual global strain parameters and EF. This is in agreement with previous studies (30–33), confirming that GCS and GLS are metrics of the contractile function. This is particularly true when considering endocardial borders, as already described with several SI techniques (34–38). Indeed, EF calculation is based on endocardial displacements, which are more directly related to endocardial deformations. However, LVEF is imputable to both longitudinal shortening and transversal thickening; therefore, the correlation of LVEF with a single strain component is necessarily, incomplete, whereas each component can reflect different aspects of systolic function or dysfunction (9). Accordingly, the higher correlation between EF^(th) (including GCS and GLS) and the SAX-EF (result of a volumetric approach of EF) is a strong argument to systematically integrate both for a better understanding of systolic function (3, 16). Although the different layers across the myocardial thickness work in concert and cannot be separated, GLS expresses more the contraction of subendocardial fibers, the first to be impacted in many pathological processes, while GCS corresponds more to medial and subepicardial fibers that are generally impacted at a later stage of the disease (14, 16, 39). Identifying these myocardial transmural patterns can be a way to a better understanding of pathological or adaptative myocardial remodeling.

Nowadays, cardiologists and radiologists are always in search for integrated tools that could give more hemodynamic and functional information in a shorter time. Artificial intelligence enables constant progress especially in cardiology (40). The integration of the EF^(th) (including GCS and GLS) in CMR software's could be a way to give access to mechanical and volumetric information in a single click without the need for specific sequence.

Finally, FT-strain imaging provides only limited insights into segmental contraction of both ventricles. Indeed, this sequence appears to be less reliable due to significant variations that depend on the spatial resolution which depend on the cardiac frequency and the number of phases of the cine sequences (41). Among the various available techniques, the Strain-encoded sequence, has demonstrated superiority over FT in assessing segmental myocardial wall motion and has comparable performance in terms of reproducibility (42). Technical advance has made possible its acquisition within a breath hold (fast-SENC) making this sequence equally usable in clinical practice than FT (43). Furthermore, it has been established as a reliable prognostic indicator in heart failure, coronary artery disease, transplant vasculopathy (44–46) and enhance the performance of CMR stress testing (47). On the other hand, inward displacement emerges as a promising and complementary approach to FT strain, given its application in cine sequences. However, like any parameter, its efficacy needs to be thoroughly studied across different pathological conditions before being implemented in clinical practice (48, 49).

Acquisition of CMR strain imaging in a heterogeneous population is feasible and reproducible—with an exception for the SAX radial strain—across a wide range of LV morphologies. These measurements have a good correlation with the gold-standard SAX derived LVEF. Finally, the reported reliability of the theoretical relationship between global strain and volumetric changes supports their conceptual integration and paves the way for a more comprehensive assessment of myocardial pathologies but still warrants further studies.