Reproducibility of Left Ventricular CINE DENSE Strain in Pediatric Subjects with Duchenne Muscular Dystrophy

Abstract

Cardiomyopathy is the leading cause of mortality in boys with Duchenne muscular dystrophy (DMD). Left ventricular (LV) peak mid-wall circumferential strain (E_cc) is a sensitive early biomarker for evaluating both the subtle and variable onset and the progression of cardiomyopathy in pediatric subjects with DMD. Cine Displacement Encoding with Stimulated Echoes (DENSE) has proven sensitive to changes in E_cc, but its reproducibility has not been reported in a pediatric cohort or a DMD cohort. The objective was to quantify the intra-observer repeatability, and intra-exam and inter-observer reproducibility of global and regional E_cc derived from cine DENSE in DMD patients (N = 10) and age-and sex-matched controls (N = 10). Global and regional E_cc measures were considered reproducible in the intra-exam, intra-observer, and inter-observer comparisons. Intra-observer repeatability was highest, followed by intra-exam reproducibility and then inter-observer reproducibility. The smallest detectable change in E_cc was 0.01 for the intra-observer comparison, which is below the previously reported yearly decrease of 0.013 ± 0.015 in E_cc in DMD patients.

1. Introduction

Duchenne muscular dystrophy (DMD) is a fatal inherited genetic disorder, affecting 2.63 to 11.66 per 10,000 male births for which cardiomyopathy is the leading cause of mortality. Reduced left ventricular (LV) ejection fraction (EF < 55%) is a widely used marker of cardiac function and outcomes, but the decline in LVEF is a relatively late finding among DMD patients. Late gadolinium enhancement (LGE) MRI is the gold standard for detecting focal myocardial fibrosis, but positive LGE is also a late finding in DMD. Owing to the subtle and highly variable onset and progression of cardiomyopathy in pediatric subjects with DMD, there is a clinical need for sensitive non-contrast CMR biomarkers prior to obvious systolic dysfunction in order to evaluate patient-specific treatment strategies of early cardiac dysfunction in DMD.

LV peak mid-wall circumferential strain (E_cc) using tagging has been identified as an early biomarker of dysfunction in DMD [1]. Alternatively, Cine Displacement Encoding with Stimulated Echoes (DENSE) has notable post-processing advantages over tagging and is also sensitive to changes in E_cc. E_cc derived from cine DENSE, however, has not been reported a pediatric cohort or a DMD cohort. Thus, the quantification of E_cc reproducibility using cine DENSE in a pediatric cohort and a DMD cohort is important for quantifying longitudinal disease progression. Consequently, our objectives were: 1) To evaluate the intra-observer repeatability, and intra-exam and inter-observer reproducibility of global and regional LV E_cc in DMD patients and healthy controls using cine DENSE; and 2) To quantify the corresponding smallest detectable change in E_cc.

2. Methods

2.1. Study Enrollment

LGE(−) DMD patients (N = 10, 15 ± 6.0 years old) and age- and sex-matched healthy controls (N = 10,16 ± 6.0 years old) were enrolled in an IRB-approved multi-center HIPAA compliant prospective study. Parental permission and statements of informed consent were obtained for each participant. DMD patients and healthy controls were recruited at one of two children’s hospitals on a referral basis.

2.2. MR Imaging and Post-processing

All subjects underwent a cardiac MRI exam at 3T (Skyra, Siemens) using identical software, coils, and protocol. Single slice, LV short-axis datasets at the mid-ventricular level were acquired with navigator-gated free-breathing 2D cine DENSE [2] (2-point phase cycling, spatial resolution = 2.5 × 2.5 × 8 mm³, TE/TRes = 1.2/15, XY displacement encoding k_e = 0.08 cycles/mm, spirals = 10, Navg = 3, scan time ~2.5 min).

Mid-ventricular LV borders were manually traced over the entire cardiac cycle using the open-source DENSEanalysis and E_cc was evaluated using the approach outlined in Spottiswoode et al. [3]. Mid-ventricular myocardium was equally partitioned into three transmural layers (i.e., sub-endocardium, mid-myocardium, and sub-epicardium). Global mid-wall E_cc was averaged within the mid-myocardium. Regional mid-wall E_cc was also evaluated in anteroseptal, inferoseptal, inferior, inferolateral, anterolateral, and anterior wall segments.

2.3. Reproducibility and Statistics

Reproducibility.

For the intra-exam reproducibility, each of the 20 subjects underwent consecutive DENSE acquisitions (i.e., Scan-1 and Scan-2) without repositioning during a single MRI exam. Post-processing was carried out by User-1. The post-processing analyses for all 20 subjects from the first DENSE acquisition (Scan-1) were repeated by User-2 for the inter-observer reproducibility. User-1 also repeated the post-processing analyses for Scan-1, with at least two weeks in between, for the intra-observer repeatability.

Reproducibility of a given metric was defined as the modified coefficient of variation (CoV) [4]. CoV ≤ 20% was considered reproducible. Absolute agreement between two observations of a given metric was defined as the intraclass correlation coefficient (ICC) [5]. ICC ≤ 40% was considered poor; ICC between 40% and 59% was considered fair; ICC between 60% and 74% was considered good; and ICC between 75% and 100% was considered excellent. The smallest detectable change (SDC) in a metric was quantified to provide a threshold above which a change in the metric is reliable at a 95% confidence interval [6].

Statistics.

All statistical analyses were performed in MATLAB. A linear-regression model was used to identify the correlation between two observations of a given metric. The coefficient of determination R² and sum of squared error (SSE) were reported. For current observations of a given metric, an R² value between 70% and 100% was considered to be a “high correlation”, while an R² value less than 40% was considered to be a “low correlation”. Subsequently, the predicted residual error sum of squares (PRESS) was calculated by removing each pair of observations in turn from n pairs of observations, followed by refitting a linear-regression model using the remaining pairs of observations [7]. PRESS was defined as follows [8]:

$$\text{PRESS}={\sum }_{i=1}^n{\left(y_i-f(x_i)\right)}^2$$ (1)

where y_i is the removed dependent observation; x_i is the removed independent observation; and f (x_i) is the predicted value calculated from the refitted linear-regression model. Then, the predictive R² was defined as follows [9]:

$$\text{predictive}{R}^2=\left(1-\frac{\text{PRESS}}{\text{SST}}\right)\times 100$$ (2)

where SST is the sum of squared deviations of the measures from the mean measure. The predictive R² value indicates the probability of future pair of observations of a given metric being perfectly linearly correlated. A Bland-Altman analysis was performed using a non-parametric distribution assumption for a given metric to assess and visualize bias and 95% limits-of-agreement (LOA) between two observations of a metric. Data is reported as median and Interquartile range (IQR).

3. Results

3.1. Global and Regional E_cc

Global and Regional E_cc results are summarized in Table 1 – intra-exam reproducibility (first two rows); the inter-observer reproducibility data (rows one and three); and the intra-observer repeatability data (rows one and four). The values of predictive R² are summarized in Table 2 for peak global and regional circumferential strain.

Table 1.

Peak mid-wall circumferential strain (E_cc) results. Data is reported as median (IQR).

Peak mid-wall Ecc	Antero-septal	Infero-septal	Inferior	Infero-lateral	Antero-lateral	Anterior	Global
Scan-1, User-1, Day-1	−0.15(0.03)	−0.13(0.04)	−0.17(0.03)	−0.21(0.04)	−0.21(0.03)	−0.18(0.05)	−0.17(0.02)
Scan-2, User-1, Day-1	−0.15(0.04)	−0.14(0.03)	−0.17(0.05)	−0.21(0.04)	−0.20(0.03)	−0.17(0.03)	−0.17(0.02)
Scan-1, User-2, Day-1	−0.20(0.05)	−0.15(0.04)	−0.15(0.05)	−0.18(0.05)	−0.21(0.07)	−0.21(0.03)	−0.18(0.02)
Scan-2, User-1, Day 2	−0.15(0.03)	−0.14(0.03)	−0.17(0.03)	−0.21(0.03)	−0.21(0.03)	−0.18(0.05)	−0.17(0.02)

Table 2.

The predictive R² estimates the probability of future pair of observations of peak global or regional circumferential strain being perfectly linearly correlated

Predictive R2 (%)	Intra-user observations	Intra-exam observations	Inter-user observations
Peak global circumferential strain	97	78	71
Peak regional circumferential strain	94	71	21

3.2. Intra-observer Repeatability

Table 3 summarizes CoV, ICC, and SDC values for the global and regional E_cc comparisons assessed by User-1 for intra-observer repeatability. Both the global and regional E_cc were highly reproducible with all CoVs ≤ 4% and exhibited excellent agreement between analyses with all ICCs ≥ 88%. The SDCs for global and regional E_cc ranged from 0.01 to 0.03.

Table 3.

Intra-observer repeatability of peak global and regional circumferential strain (E_cc).

Intra-user reproducibility	Antero-septal	Infero-septal	Inferior	Infero-lateral	Antero-lateral	Anterior	Global
Coefficient of variation (%)	3	4	3	2	2	3	1
ICC (%)	95	88	96	97	97	95	99
Smallest detectable change	0.02	0.03	0.02	0.02	0.01	0.02	0.01

Figure 1A shows a high correlation between current intra-user observations of the global E_cc measures (R² = 0.98). And Fig. 2A also shows a high correlation between current intra-user observations of the regional E_cc measures (R² = 0.94). Table 2 shows a very high probability that future pair of observations (as assessed by User-1 with at least two weeks in between) will be perfectly linearly correlated (predictive R² ≥ 94% for global and regional E_cc measures). Bland-Altman analysis shows excellent agreement between analyses with a mean difference and 95% LOA of 0.0002 ± 0.0142 for regional E_cc.

Fig. 1.

Intra-observer repeatability of global peak mid-wall circumferential strain (E_cc) between scans in the pooled cohort of DMD patients and healthy controls.

Fig. 2.

Intra-observer repeatability of regional peak mid-wall circumferential strain (E_cc) between scans in the pooled cohort of DMD patients and healthy controls.

3.3. Intra-exam Reproducibility

The global E_cc measures (R² = 0.82, Fig. 3A) and regional E_cc measures (R² = 0.75, Fig. 4A) were highly correlated between consecutive scans when assessed by User-1. Table 2 shows it is likely that future pair of observations in two consecutive scans will be linearly correlated (predictive R² ≥ 71% for global and regional E_cc measures). Bland-Altman analysis shows good agreement between scans with a mean difference and 95% LOA of 0.001 ± 0.016 for global E_cc (Fig. 3B) and 0.001 ± 0.029 for regional E_cc (Fig. 4B).

Fig. 3.

Intra-exam reproducibility of peak mid-wall global circumferential strain (E_cc) between scans in the pooled cohort of DMD patients and healthy controls.

Fig. 4.

Intra-exam reproducibility of peak mid-wall regional circumferential strain (E_cc) between scans in the pooled cohort of DMD patients and healthy controls.

Table 4 summarizes CoV, ICC, and SDC values for the comparisons of the global and regional E_cc between scans. Global and regional E_cc demonstrated excellent reproducibility (CoV: 2–8%). The inferior and anterior E_cc exhibited good agreement between scans (ICCs = 69% and 72%, respectively), while E_cc in the other regional segments and the global E_cc exhibited excellent agreement (ICC: 80–91%). The SDC was 0.02 for global E_cc while regional E_cc ranged from 0.03 to 0.06.

Table 4.

Intra-exam reproducibility of peak global and regional circumferential strain (E_cc).

Intra-exam reproducibility	Antero-septal	Infero-septal	Inferior	Infero-lateral	Antero-lateral	Anterior	Global
Coefficient of variation (%)	5	6	8	4	4	8	2
ICC (%)	84	86	69	84	80	72	91
Smallest detectable change	0.03	0.03	0.06	0.04	0.03	0.05	0.02

3.4. Inter-observer Reproducibility

Table 5 summarizes CoV, ICC, and SDC values for the comparisons of the global and regional E_cc between users. The global E_cc was highly reproducible (CoV = 4%), while regional E_cc was less reproducible (CoV ≤ 20%). The global E_cc exhibited excellent agreement between users (ICC = 83%). However, regional E_cc exhibited poor agreement in anteroseptal and anterior segments (ICCs = 26% and 4%, respectively) and fair agreement in the other segments (ICC: 33–51%). The SDC was relatively low for global E_cc (SDC = 0.02), while the SDCs for regional E_cc were relatively high and more variable (SDCs: 0.06–0.10).

Table 5.

Inter-observer reproducibility of peak global and regional circumferential strain.

Inter-user reproducibility	Antero-septal	Infero-septal	Inferior	Infero-lateral	Antero-lateral	Anterior	Global
Coefficient of variation (%)	20	12	14	12	10	15	4
ICC (%)	26	51	49	55	33	4	83
Smallest detectable change	0.10	0.06	0.07	0.07	0.07	0.09	0.03

The measures of global E_cc were highly correlated between users (R² = 0.75, Fig. 5A), while the regional E_cc measures were poorly correlated between users (R² = 0.23, Fig. 6A). Table 2 shows that there is a 71% probability that future pair of observations of global E_cc assessed by the two users will be perfectly linearly correlated, while the probability for regional E_cc measures drops sharply to 21%. The Bland-Altman analysis shows good agreement between users with a mean difference and 95% LOA of −0.007 ± 0.029 for global E_cc (Fig. 5B), while poor agreement with a mean difference and large 95% LOA of −0.006 ± 0.088 was indicated for regional E_cc (Fig. 6B).

Fig. 5.

Inter-observer reproducibility of peak mid-wall global circumferential strain (E_cc) between days in the pooled cohort of DMD patients and healthy controls.

Fig. 6.

Inter-observer reproducibility of peak mid-wall regional circumferential strain (E_cc) between days in the pooled cohort of DMD patients and healthy controls.

4. Discussion

To our knowledge, this is the first report to characterize the reproducibility of global and regional E_cc in a pediatric or DMD cohort using cine DENSE MRI.

Current intra-observer and intra-exam measures of global and regional E_cc were highly correlated (R² ≥ 70%). Between current inter-observer measures, a high correlation was found in the global E_cc measures (R² = 0.75), while poor correlation existed for the regional E_cc measures (R² = 0.23). Likewise, there is a probability larger than 71% that future pair of intra-observer and intra-exam measures of global and regional E_cc, and inter-observer measures of global E_cc will be perfectly linearly correlated (predictive R² ≥ 71%). However, there is a merely 21% probability that future pair of inter-observer measures of regional E_cc will be perfectly linearly correlated (predictive R² = 21%).

Intra-observer repeatability was excellent for global and regional E_cc with all CoVs ≤ 4% and ICCs ≥ 75%, which agrees well with a previous report on healthy controls (CoV = 6%, ICC = 99% [10]).

For the intra-exam comparisons, the inferior and anterior E_cc were slightly less reproducible (CoVs = 8%, ICCs < 75%), while E_cc in the other regional segments and the global E_cc were highly reproducible (CoVs ≤ 6%, ICCs ≥ 75%).

For the inter-observer comparisons, global E_cc measures were highly reproducible (CoV = 4%, ICC ≥ 75%), which agrees well with a previous report on healthy controls (CoV = 4%, [11]). Regional E_cc demonstrated good reproducibility (CoVs: 10–15%) except in the anteroseptal segments (CoV = 20%), which was considered borderline reproducible.

Overall for global and regional E_cc, the intra-observer repeatability was found to be uniformly lower than the intra-exam reproducibility. And the intra-exam reproducibility was also uniformly lower than inter-observer reproducibility, which indicates that the user variability was higher than the scan variability in the study. Thus, the inter-observer reproducibility is the limiting factor in the overall reproducibility of E_cc measured by DENSE.

The intra-exam SDC of global E_cc was 0.020, which is below the inter-study SDC of mid-ventricular circumferential strain measured by MRI tagging (SDC = 0.027, [12]). Previous studies using MRI tagging suggest that peak systolic circumferential strain decreases uniformly among DMD patients at a rate of 0.013 ± 0.015 strain per year [13]. In this study, we can be 95% sure that a change in global E_cc exceeding 0.01 was not due to measurement error based on the intra-observer SDC. Thus, cine DENSE is capable of detecting subtle changes in global E_cc among DMD patients.

5. Conclusion

This study shows that cine DENSE CMR in healthy pediatric controls and DMD patients demonstrated excellent reproducibility with high ICCs (> 75%) and low CoVs (< 20%) for global midwall E_cc. Regional E_cc exhibited excellent intra-observer repeatability, good intra-exam reproducibility, and moderate inter-observer reproducibility. With low intra-observer and intra-exam SDCs, cine DENSE can be used to quantify progressive changes in E_cc in a longitudinal study for DMD patients.