Typically, DENSE acquisition is carried out in three short axial segments of the LV: the base, the mid-ventricle and near the apex. Imaging can also be performed in the longitudinal four-chamber plane. This imaging technique is not limited to the LV but can also be used for assessing the right ventricle (RV). The DENSE technique uses a combination of radiofrequency pulses and gradients to encode myocardial displacement into the phase of the magnetic resonance imaging (MRI) signal (Fig. 1). The first pulses and the first gradient are applied at the end-diastole. After the first radio frequency pulse, a spatial magnetic field gradient pulse (modulation; M) imparts a location-dependent phase shift to the stimulated echo. A strong pulse (crusher; C) removes the transversal magnetization components. Second, another gradient pulse (demodulation; DM) removes the initial phase shift. The residual phase shift reflects tissue displacement between the two pulses [30]. Each phase image measures tissue displacement in a single direction only. For this reason, displacement requires two recordings to measure both horizontal and vertical directions [8]. This information is then further acquired and processed. The phase data is unwrapped, and spatial smoothing is applied to the displacement data to reduce noise and improve the accuracy of the displacement estimates [31]. Finally, temporal fitting is performed to generate continuous displacement fields over time, which can be used to calculate strain and other parameters of interest [31]. A recently published paper introduced an approach for calculating Lagrangian tissue displacement and strain in cine DENSE MRI. This method utilizes a regularized spatiotemporal least squares technique, providing a novel way to analyze and measure cardiac tissue movement and deformation [32].

Fig. 1.

Simplified illustration of displacement encoding with stimulated echoes (DENSE) acquisition. After the radio frequency pulses, the modulation (M) pulse and its counter pulse (demodulation; DM) are induced. The pulses are the same size, and they are sent in the direction in which the tissue movement is to be measured. A strong pulse (crusher; C) is given between them, which removes the transversal magnetization components. The movement of the tissue can then be evaluated with the help of the difference between the modulation pulse and the counter pulse. TE/2, echo time; TM, mixing time; DA, data acquisition.

The DENSE dyssynchrony analysis method uses information from myocardial strains to provide insights into the changes in myocardial tissue dimension during the cardiac cycle [33]. By measuring strains, it is possible to assess the variability of contraction over time between different areas of the heart, as well as to determine the degree of synchronicity. This can be presented in milliseconds (ms) or relative to the duration of the cardiac cycle (% of RR interval [the distance between two consecutive R waves]). An example of dyssynchrony analysis is presented in Fig. 2 (Ref. [7]).

Fig. 2.

The DENSE dyssynchrony analysis method. Displacement encoding with stimulated echoes (DENSE): magnitude (A) and phase (E) images of a patient with heart failure and left bundle branch block. The corresponding displacement map (B), Ecc (circumferential strain) map (C), and segmental Ecc-time curve (D). Modified from Auger et al. 2022 [7], license: CC BY 4.0.

The regional heterogeneity of strains in circumferential (circumferential uniformity ratio estimate [CURE]), longitudinal (longitudinal uniformity ratio estimate [LURE]) or radial (radial uniformity ratio estimate [RURE]) planes are the most-used methods for dyssynchrony evaluation [34]. The uniformity ratio is determined from several locations in the selected plane around each imaging segment or slice and plotted versus spatial position for each time frame. The more oscillatory the plot, the greater the dyssynchrony among the segments. Plots are further subjected to Fourier analysis, and the results are averaged over space and time [35]. The obtained values are scaled between 0 (indicating dyssynchrony) and 1 (indicating synchrony) [34]. CURE with singular value decomposition (CURE-SVD) offers an advantage over standard CURE, since it does not require the selection of specific cardiac phases [36]. This is particularly important in patients with HF and dyssynchronous LV, where there is often ongoing contraction in some segments of the myocardium and simultaneous stretch in others. Yet, a previous study showed that the correlation between CURE and CURE-SVD is high (r = 0.90, p 0.0001) [37]. Different DENSE dyssynchrony measurement methods are listed in Table 1 (Ref. [34, 36, 38, 39, 40, 41]).

The dyssynchrony index (DI) is another parameter that can be calculated by measuring the timing of contraction for each segment throughout the heart and establishing a person-specific reference strain [38]. This approach allows for a more detailed assessment of the degree of dyssynchrony in the myocardium [38]. In this method, a temporal offset (TO) is measured in ms for each segmental strain curve relative to the LV reference curve. These TOs are then mapped onto a bullseye diagram and smoothed spatially using a cubic spline. The DI is then calculated as the standard deviation of the TOs. The DI can be calculated intra-ventricularly, inter-ventricularly or separately for specific regions of the heart, such as the septum and inferolateral wall. Another parameter used in DENSE studies is the time to the onset of circumferential shortening (TOS) [39]. The TOS is the point in time when the downslope of the regional circumferential strain curve (Ecc) begins. Specifically, TOS corresponds to the moment when the rate of change of Ecc with respect to time (dEcc/dt) transitions from a value near zero to a negative value after the system detects the ECG R-wave [39]. Additionally, TOS can be utilized to assess the diversity of the contraction and identify regions that activate later. For instance, visual representations, such as bull’s-eye plots of TOS maps, can effectively illustrate synchronicity of LV contraction [39].

Myocardial MRI tagging is used to analyze the motion and deformation of the myocardium during the cardiac cycle. This technique involves the use of radiofrequency pulses to create a grid-like pattern of dark and light lines (tags) within the myocardium [1]. Harmonic phase (HARP) is a more advanced analysis technique capable of extracting and processing motion information from tagged CMR. It is obtained by calculating both motions from the horizontal direction and vertical direction, resulting in a 2D vector field [42]. Strain-encoded (SENC) MRI uses tag surfaces parallel to the image plane, combined with out-of-plane phase-encoding gradients along the slice-selected direction, to measure myocardial strain [43]. Feature-tracking (FT) analysis derives information from routine cine images by following ‘features’ across multiple images based on pattern-matching techniques without the need for additional imaging [44]. This makes the technique more efficient and less time consuming. However, the technique also has limitations, including the potential for inaccuracies due to noise and the variability in tracking performance between the different software packages [45, 46]. Moreover, FT relies on the quantification of local changes in signal intensity, which may fail when quantifying motion parallel to or inside the myocardium. Additionally, it only captures in-plane displacement, which may limit its accuracy. Compared to DENSE, FT has been shown to overestimate dyssynchrony (bias: 26 ms, p 0.001, coefficient of variation [CoV]: 76%) [47].

Echocardiography is a widely used method for assessing LV volumes and ejection fraction (EF), since it displays good time and spatial resolution. Moreover, echocardiography has high availability and relatively low cost. Yet, echocardiography has some limitations, including its relatively high inter-observer variability and, in some cases, poor visibility, particularly on the right side of the heart [19]. However, the consistency of echocardiographic parameters used to measure dyssynchrony has been modest, according to the Predictors of Response to CRT Trial (PROSPECT), which is one of the largest and most widely referenced investigations in this field [19].

Dyssynchrony assessment with single-photon emission computerized tomography (SPECT) is based on regional changes in LV counts during the cardiac cycle. Phase analysis quantifies the heterogeneity of contraction onset times in various regions of the LV and displays the results as a histogram. The wider the histogram, the more substantial the dyssynchrony. Dyssynchrony can also be measured using other nuclear imaging techniques such as multi-gated acquisition radionuclide angiography and gated blood-pool SPECT [48]. However, there are several disadvantages to this method, including radiation and relatively low spatial resolution. Also, RV or atrial analysis is not possible with this technique.

RV function is an important indicator of various cardiopulmonary diseases [49]. However, the complex anatomy and thin wall of the RV make it challenging to assess its function using conventional imaging techniques [50]. Echocardiography is widely used for assessing RV function but has limitations due to the retrosternal position of the RV. Three-dimensional echocardiography and myocardial deformation imaging may help overcome these limitations. Nuclear imaging techniques can estimate LV dyssynchrony but are unsuitable for assessing RV function. MRI offers excellent tissue contrast and multi-slice imaging capabilities, making it useful for assessing RV mechanics. Yet, only a few MRI studies have investigated RV synchronicity and contraction patterns.

CRT is a treatment for advanced HF that involves implanting a device to resynchronize LV contractions. However, nearly a of third patient selected utilizing the traditional inclusion criteria for CRT (EF $\le$35%, QRS duration $\ge$120 ms, left bundle branch block, with moderate or severe HF symptoms) have an inadequate response [18, 19]. It has been shown that dyssynchrony measurement can help in CRT patient selection [51, 52, 53, 54] and can also be useful for monitoring disease progression and response to treatment in patients with HF [55]. Moreover, measuring LV dyssynchrony can enhance our understanding of the underlying pathophysiology of HF [56].

Four electronic databases were searched: PubMed (US National Library of Medicine, http://www.ncbi.nlm.nih.gov/pubmed), Scopus (Elsevier, http://www.scopus.com/), Web of Science (Thomson Reuters, http://apps.webofknowledge.com/) and Cochrane (https://www.cochrane.org/). These databases are widely recognized in the scientific community as reliable sources of scholarly publications and were chosen for their comprehensive coverage of medical research literature. By utilizing multiple databases, this study aimed to ensure a thorough and comprehensive search for relevant papers.

To identify relevant studies, the following search criteria were used: (DENSE OR ‘displacement encoding with stimulated echoes’ OR CURE) AND (dyssynchrony* OR asynchron* OR synchron*) AND (MRI OR ‘magnetic resonance’ OR CMR). The search was conducted for articles published from January 2000 through January 2023 and included the use of relevant keywords and their variations.

First, the researchers excluded duplications. Next, they analyzed the titles and abstracts of various studies and only included those that reported results considering CMR related imaging. The studies had to meet specific inclusion criteria, such as being clinical research or technical development studies and having been published in English. The researchers excluded reviews, general overview articles and case reports. The final articles were those that considered the DENSE technique and especially dyssynchrony imaging. During the selection process, two readers worked together to decide which articles to include in the analysis. The researchers extracted information, such as authors, year of publication, sample size and diagnostic modality, from each study.

We identified only one study that provided DENSE values for a healthy human population. The study from Suever et al. [41], involved 50 healthy individuals and measured global cardiac mechanics for both LV and RV using a 3 Tesla (Tim Trio) machine. The timing of contraction was assessed across the ventricles by calculating the mechanical activation delay in ms for each segment in relation to a patient-specific reference curve and further converted from ms to percent of the cardiac cycle delay time to provide DI. The mean global 3D DENSE values for LV were 25.0 $\pm$ 6.9 ms and DI 3.4 $\pm$ 1.0%. For RV, the mean global 3D DENSE values were 23.3 $\pm$ 8.3 ms and DI 3.1 $\pm$ 1.1%. Inter-ventricular values were –0.7 $\pm$ 10.6 ms and DI –0.0 $\pm$1.5% respectively.

Previous studies have demonstrated that the reproducibility of DENSE strain is superior to that of tagging [6]. The reproducibility of DENSE dyssynchrony parameters has been evaluated in several studies. Suever et al. [41] also studied the inter-observer variability of DI from 10 healthy individuals and the inter-study reproducibility from 6 healthy individuals using the CoV. For LV DI, the CoV was 6%, and for RV DI, the CoV was 10%. The inter-study reproducibility for LV DI was 12% and for RV DI 11%. Inter-observer variability was also reported for DI in the rTOF population. For the LV, the CoV was 9% and the intra-class correlation (ICC) 0.76 (95% CI 0.32–0.93); for RV DI, the CoV was also 9%, and the ICC was 0.84 (95% CI 0.51–0.96); and for inter-ventricular DI, the CoV was 7% and the ICC 0.90 (95% CI 0.66–0.97) [38]. Haggerty et al. [57] studied reproducibility on five healthy mice (C57BL/6) and four mice with impaired cardiac function (diet-induced obesity), which were imaged with a 7T ClinScan MR system. CURE and RURE were used as indices of synchrony. The inter-observer variability, expressed as (CoV), was 1% for CURE and 3% for RURE. Similarly, the inter-study variability was 1% for CURE and 3% for RURE.

We found seven studies on DENSE dyssynchrony analysis related to CRT response. However, only one provided optimal cut-off values with accuracy analysis predicting patient outcome. Bilchick et al. [40] showed that patients who have CURE below 0.70, a delay in electrical contraction at the LV lead position and no scarring in that location are more likely to experience the benefits of CRT. The study consisted of 75 patients and 40 (53%) patients identified with positive CRT response. Multivariable logistic modeling accurately identified CRT responders (area under the curve: 0.95 [p 0.0001]) based on CURE (odds ratio [OR]: 2.59/0.1 decrease), delayed circumferential contraction onset at left ventricular lead position (LVLP) (OR: 6.55), absent LVLP scar (OR: 14.9), and the time from QRS onset to LVLP electrogram (OR: 1.31/10 ms increase). Patients with a favorable CMR profile, characterized by a CURE less than 0.70, no LVLP scar, and delayed electrical conduction onset at LVLP, demonstrate better overall survival compared to patients without dyssynchrony as determined by a CURE score of 0.70 or higher [40]. In that study optimal cut-off value was for CURE 0.70 with sensitivity or 100% and specificity of 65.7%. Ramachandran et al. [36] showed that patients with positive CRT response had statistically more LV dyssynchrony and lower CURE (0.38 vs. 0.75, p 0.0001) and CURE-SVD (0.45 vs. 0.84, p 0.0001) compared to non-responders. Another study compared DENSE to other CMR dyssynchrony methods and showed that CURE-SVD measured with DENSE had a stronger correlation with CRT response (r = –0.57, p 0.0001) than CURE-SVD with FT (r = –0.28, p = 0.004) [37].

In a recently published prospective study of 200 patients, the influence of sex on CRT response was evaluated. The study found that females had more mechanical dyssynchrony before CRT installation compared to males, as measured by the CURE-SVD parameter (0.52 vs. 0.61, p = 0.04) [58]. Additionally, female patients had a smaller LV end-diastolic volume index and a lower frequency of both late gadolinium enhancement and ischemic cardiomyopathy. To further analyze the data, patients were categorized into four groups based on sex and cardiomyopathy type (ischemic or non-ischemic). Female patients with non-ischemic cardiomyopathy had the lowest pre-CRT CURE-SVD, indicating more dyssynchrony (p = 0.003), but also had the most favorable response for LV function. Moreover, female patients had better three-year survival compared to men. Another study conducted by the same research group and utilizing the same dataset delved into the application of machine learning (ML) for predicting CRT response and long-term survival [59]. The study examined the associations of 39 baseline features, and ML-generated response clusters were evaluated, with cross-validation assessing the associations of clusters with four-year survival. The study found that lower CURE-SVD values (indicative of more dyssynchrony) were associated with greater CRT response. Additionally, the top three pre-CRT predictors were CURE-SVD, pre-CRT B-type natriuretic peptide and pre-CRT peak VO${}_2$. The resulting model was able to provide reassurance to approximately 62% of patients that their four-year survival was expected to be favorable, while also identifying 16% of patients who would benefit from further evaluation or advanced HF therapies after CRT.

Bilchick et al. [37] studied the prognostic value of DENSE for CRT patients by combining the Seattle Heart Failure Model (SHFM) and DENSE CURE-SVD circumferential strain dyssynchrony parameter. The SHFM score provides estimates for one-, two- and three-year survival with the use of the clinical, pharmacological, device and laboratory characteristics [68]. In a cohort of 100 patients who were followed a median of 5.3 years, CURE-SVD and SHFM were independently associated with the primary endpoint of death, heart transplantation, LV assist device and appropriate implantable cardioverter-defibrillator therapies (SHFM: hazard ratio (HR) 1.47/standard deviation (SD), 95% CI 1.06–2.03, p = 0.02; CURE-SVD: HR 1.54/SD, 95% CI 1.12–2.11, p = 0.009) [37].

Previous studies have suggested that using cardiac imaging could be advantageous in guiding the placement of the CRT lead [69]. Auger et al. [39] used TOS to evaluate the synchronicity and the latest contraction site of the LV. They showed that TOS was associated with an improvement in LV reverse remodeling with CRT.

Suever et al. [41] determined RV geometry from 50 healthy individuals (age: 26 $\pm$ 8 years, 46% male) with no history of cardiovascular disease by using DENSE. The timing of contraction was assessed across the entire RV by calculating the mechanical activation delay in ms for each segment in relation to a patient-specific reference curve and further converted from ms to percent of the cardiac cycle delay time to provide DI. For the RV, the mean 3D DENSE values were 23.3 $\pm$ 8.3 ms and DI 3.1 $\pm$ 1.1.

Haggerty et al. [60] studied 40 patients with rTOF. They found a significant association between LV dyssynchrony and myocardial fibrosis in their study of rTOF patients using spiral cine DENSE and modified Look-Locker inversion recovery (MOLLI) T1-mapping. Specifically, they found that extracellular volume fraction, a measure of myocardial fibrosis, was positively associated with a log-adjusted DI ($\beta$ = 0.67), suggesting a correlation between cardiac fibrosis and mechanical dyssynchrony.

Kramer et al. [61] randomized ten 12-week-old C57BL/6 J mice to a high-fat (60% of calories from fat) or low-fat (10% of calories from fat) diet. After five months on the diet, mice were imaged with a 7 T Bruker ClinScan using a cine DENSE protocol. To evaluate the myocardial contraction, LV strains were calculated and used to quantify LV synchrony using CURE and RURE indices [61]. Kramer et al. [61] found that obese mice had a 15% increase in LV mass compared to the control mice. Interestingly, there was no difference in EF between the groups (p = 0.056). The synchrony of contraction in the LV was decreased in obese mice; RURE was 0.95 $\pm$ 0.02 in the control mice and 0.91 $\pm$ 0.03 in the obese mice (p = 0.032). CURE did not differ across the two groups of mice (p = 0.151).