T₁-Mapping and extracellular volume estimates in pediatric subjects with Duchenne muscular dystrophy and healthy controls at 3T

This two-center prospective CMR study was approved by both Institutional Review Boards and completed between January 2017 and January 2020. We obtained parental consent and child assent for all study participants under the age of 18 years. Boys with DMD were recruited from one of two children’s hospitals on a referral basis from two counties with large urban populations. DMD diagnosis was confirmed by genetic testing to identify the presence of a dystrophin mutation. Boys with DMD did not require respiratory support and were not enrolled in any therapeutic clinical trial at the time of the study. Healthy participants were recruited from the surrounding communities and had no history of cardiovascular disease. Site-A enrolled 13 healthy controls and 21 boys with DMD. Site-B enrolled 7 healthy controls and 7 boys with DMD. In total, 28 boys with DMD and 20 sex-matched healthy controls were enrolled. Table 1 displays the demographic information for the two groups and a summary of medications taken by boys with DMD at the time of the CMR exam.

After providing informed consent, all 45 participants underwent 3 T CMR (Skyra, Siemens Healthineers, Erlangen, Germany) at each site, using identical software, coils, and scan protocol. The CMR exam included standard functional imaging using a high spatial and temporal resolution, free-breathing retrospectively binned balanced steady state free precession (bSSFP) cine sequence [19, 20] with the following acquisition parameters: 40º flip angle, 6/8 partial Fourier and rate-4 parallel imaging, matrix size 192 × 144, pixel size 1.9 mm × 1.9 mm, slice thickness 8 mm, bandwidth (BW) 930 Hz/Px, TE/TR 1.2 ms/2.4 ms echo spacing, and a temporal resolution of 64.4 ms. Breath-held bSSFP cine was used in six study participants when the free-breathing sequence was unavailable. Parameters for this acquisition were: 58º flip angle, rate-3 parallel imaging, matrix size 256 × 192, pixel size 1.6 mm × 1.6 mm, slice thickness 6 mm, BW 977 Hz/Px, TE/TR 1.4 ms/3.3 ms echo spacing, and a temporal resolution of 32.5 ms. All cine imaging spanned the entire LV from base to apex using short-axis slices.

Myocardial T₁ measurements were acquired using a modified Look-Locker inversion recovery (MOLLI) sequence with MOtion Correction (MOCO) [11] in a single mid-ventricular short-axis (SAx) slice and was performed with electrocardiographic (ECG)-gating and breath holding. Pre- and post-contrast T₁ mapping was acquired with a 5(3 s)3 and a 4(1 s)3(1 s)2 MOLLI scheme, respectively. Typical native T₁ imaging parameters were: non-selective inversion pulse, bSSFP single shot  readout with a 20º excitation flip angle, 7/8 partial Fourier and rate-2 parallel imaging, matrix size 192 × 132, pixel size 1.9 mm × 1.9 mm, slice thickness 8 mm, BW 1085 Hz/Px, minimum inversion time (TI) of 100 ms and incremented by 80 ms, TE/TR 1.01 ms/2.44 ms echo spacing. Typical post-contrast T₁ imaging parameters were: non-selective inversion pulse, bSSFP single shot read out with a 20º excitation flip angle, 7/8 partial Fourier and rate-2 parallel imaging, matrix size 192 × 164, pixel size 1.9 mm × 1.9 mm, slice thickness 8 mm, BW 1085 Hz/Px, minimum TI of 100 ms with 80 ms increments, TE/TR 1.01 ms/2.44 ms echo spacing. For heart rates greater than 90/min, the matrix size was decreased to 192 × 128 to mitigate any heart rate biases. Two subjects were inadvertently scanned with the 4(1 s)3(1 s)2 MOLLI scheme for their pre-contrast scan. We have found that this error does not significantly impact the group pre-contrast results. We computed the percent error between each such subject’s pre-contrast T₁ and the group pre-contrast mean and found the measurements themselves do not vary significantly from the group mean (percent error < 5% for both measurements).

Gadobenate dimeglumine contrast (Gd-BOPTA, MultiHance, Bracco Diagnostics, Milan, Italy) was administered either by hand or computer controlled injection only to boys with DMD. Eight minutes following contrast administration, LGE imaging with a free breathing motion corrected phase sensitive inversion recovery (PSIR) sequence [21] was acquired in the short axis (SAx) view spanning base to apex using the following parameters: 20º flip angle, rate-2 parallel imaging, matrix size 192 × 120, pixel size 1.4 mm × 1.4 mm, slice thickness 6 mm, BW 977 Hz/Px, TE/TR 2.01 ms/2.83 ms echo spacing, and a temporal resolution of 35.1 ms. Vertical long axis (VLA) and horizontal long axis (HLA) views were also acquired. Approximately 18 min (18 \(\pm\) 6.1 min) after contrast injection, post-contrast T₁ mapping was performed at slice locations matched to the pre-contrast acquisition. The MOCO T₁ (pre- and post-contrast) maps were generated on the scanner and later used for calculating ECV maps. All boys with DMD provided a blood sample on the day of the CMR exam for measurement of hematocrit to be used in calculating the subject specific ECV.

Two expert clinicians (either PR or AP) contoured the images, individually at their respective site, using Circle CVI42 (Circle Cardiovascular Imaging Inc., Calgary, Canada) and Medis (Medis Cardiovascular Imaging, Leiden, the Netherlands). They calculated and analyzed the following functional metrics: LV end systolic volume (LVESV) and end diastolic volume (LVEDV), ejection fraction (LVEF), and LV mass (LVM). Parameters were indexed by body surface area (BSA) to derive LVESVI, LVEDVI, and LVMI. A normal LVEF was defined as LVEF ≥ 55% [22]. Additionally, the clinicians noted the presence or absence of LGE and indicated the number of affected segments according to the American Heart Association (AHA) 17-segment model [23]. A patient with LGE presence in at least one myocardial segment was considered to be LGE positive (LGE +). If no enhancement was observed, then the subject was identified as LGE negative (LGE−). 26 LGE exams were analyzed by both clinicians with two exams excluded due to poor image quality.

Pre- and post-contrast T₁ maps were registered using a combination of two-dimensional rigid and affine image registration techniques using MATLAB (MathWorks, Natick, Massachusetts, USA) software, then combined with each DMD participant’s hematocrit to calculate an ECV map [9]. A region of interest (ROI) encompassing the LV myocardium, and two additional ROIs including a septal and lateral wall segment, were manually selected and analyzed. From each ROI, summary statistics including within-slice standard deviation (SD) were extracted for the global, septal, and lateral LV myocardial regions (Additional file 1: Figure S1). An agreement analysis of the T₁ mapping measurements between the two sites was performed to ensure the applicability of both DMD and healthy control data in the group-wise comparisons. Site-specific measurements were compared using a Wilcoxon-rank sum test.

Demographics for the boys with DMD and healthy controls were compared using a Wilcoxon rank-sum test. Following skewness and kurtosis tests for normality, group-wise comparisons of segmental and global myocardial pre-contrast T₁ were performed with a Wilcoxon rank-sum test between boys with DMD and healthy controls. The segmental and global myocardial post-contrast T₁ and ECV data were compared for two DMD sub-groups (LGE + vs. LGE-). Furthermore, the within-slice standard deviation (SD) for pre-contrast and post-contrast T₁ and ECV was evaluated in an effort to characterize differences in myocardial tissue heterogeneity between boys with DMD and healthy controls, and also between the two DMD sub-groups. After post hoc correction for multiple comparisons, a p-Value < 0.05 was considered significant. Due to the varied progression of DMD within this patient cohort and non-normal distribution of the CMR measurements, data is reported as median and interquartile range (IQR). A linear-regression analysis was used to identify initial correlations between T₁ metrics and LV function in boys with DMD and healthy controls. R² and p-Values are reported. Multiple-regression analysis was then used to test for correlations between T₁-mapping (pre- and post-contrast T₁, and ECV) measured from lateral wall segments, and global functional metrics (LVEF, LVEDVI, LVESVI, LVMI), and Age, BMI, and heart rate covariates. A binomial logistic regression classifier was analyzed using Receiver Operating Characteristic (ROC) analysis for each measured biomarker in the following distinguishing tasks: (1) healthy controls vs. DMD; (2) healthy controls vs. LGE− boys with DMD; and (3) LGE− vs. LGE + boys with DMD. Results are displayed by ROC curves and the area under the ROC curve (AUC) is reported. ROC curves and AUCs are compared for each individual biomarker. A combination (native T₁ and LVEF) is used to evaluate the discriminatory power of non-contrast biomarkers. All statistical analyses were performed in MATLAB (Mathworks).

We found four significant demographic differences between the two groups: boys with DMD had faster heart rates and were shorter, resulting in larger BMI and smaller BSA values compared to healthy controls (Table 1).

Boys with DMD had significantly reduced LVEF [49.5 (11.3) % vs 55.9 (5.8) %); p-value = 0.003] and lower LVMi [35.6 (9.8) g/m² vs 38.4 (7.8) g/m²; p-value = 0.04]. Among boys with DMD, 17 out of the 28 (61%) presented with reduced LVEF. There were no significant differences in LVEDVI, and LVESVIi between the two groups, but boys with DMD had a smaller LVEDVI and larger LVESVI compared to healthy controls. Indices of LV function are displayed in Table 2.

Nine (32%) of the DMD boys were LGE+ with at least one myocardial segment. Figure 1 shows the distribution of LGE+ segments for all nine LGE+ boys with DMD. Furthermore, all LGE+ boys had enhancement present in the mid-ventricular slice used for T₁ mapping. One significant demographic difference was observed in this group: LGE+ DMD boys had a lower heart rate [73.5 (14.5) bpm vs 96 (28.8) bpm; p = 0.01] compared to LGE− patients. For all LGE+ boys with DMD, the clinicians observed enhancement present at the lateral LV wall, but in two boys, enhancement was also present at the septal wall. LGE+ patients with DMD had a significantly larger LVEDVi [91.8 (39.6) g/m² vs. 68.2 (26.8) g/m²; p = 0.02] and LVESVI [45.5 (28.7) g/m² vs 36.4 (9.7) g/m²; p = 0.001] and lower LVEF than LGE− patients with DMD [44.8 (10.7) % vs 55.2 (10.9) %; p = 0.005].

We performed an agreement analysis between the two sites (i.e. Site-A and Site-B) and found no statistically significant differences in the measured native T₁ values between healthy controls (Fig. 2a) and boys with DMD (Fig. 2b). Similarly, no significant difference was observed in ECV measurements (Fig. 2d) between the two sites. Post-contrast T₁ measurements between the two sites, however, were found significantly different for all three regions of interest (global, septal, and lateral) as seen in Fig. 2c. Post-contrast measurements from Site-B were significantly lower than measurements from Site-A. To better understand these site specific differences, we further assessed the pre- and post-contrast blood pool T₁ measurements from all controls and boys with DMD, the blood hematocrit, and the average time after contrast injection for boys with DMD (Table 3). Significant differences were found only in the post-contrast blood pool measurements [406 (197) ms vs. 324 (100) ms; p = 0.03] between Site-A and Site-B and in the blood hematocrit measurements [44 (2.4) % vs. 40 (3.0) %; p = 0.01], respectively.

All native T₁ maps were analyzed for the healthy controls. Within the DMD cohort, three patients were unable to complete the entire CMR exam resulting in 27 (96%) native T₁, and 25 (89%) post-contrast T₁ and ECV maps analyzed. Figure 3 displays example pre-contrast and post-contrast T₁ maps (columns A-B), ECV maps (column C), and LGE images (column D) for three boys with DMD at varying stages of cardiac involvement. Guided by previous studies [24, 25], boys with DMD for this study were defined to be in the early stages of cardiac involvement if they were LGE− with normal LVEF. Mid-stage patients were defined more broadly: (1) either a patient was LGE− with reduced LVEF or; (2) also if the patient was LGE+ with normal LVEF. Advanced stage cardiac involvement was defined as LGE+ with reduced LVEF and visibly dilated LV. Tables 4 and 5 summarize the T₁ mapping results.

Compared to healthy controls, DMD subjects had significantly increased global myocardial native T₁ [1289 (56) ms vs. 1332 (60) ms; p = 0.004; Fig. 4] and significantly increased within-slice pre-contrast T₁ SD [74 (27) ms vs.100 (57) ms; p = 0.001; Table 5]. In the lateral wall, native T₁ [1348 (86) ms vs. 1277 (58) ms vs.; p = 0.001] and within-slice native T₁ SD [98 (46) ms vs. 67 (27) ms; p = 0.001; Table 4] remained significantly increased in boys with DMD compared to healthy controls. The septal myocardium showed no significant differences in native T₁ [1300 (55) ms vs 1299 (38) ms; p = 0.64; Fig. 4], nor within-slice SD [63 (31) ms vs 87 (31) ms; p = 0.06; Table 5] between healthy and DMD boys, respectively.

LGE+ boys with DMD had significantly increased native T₁ [1350 (53) ms vs 1315 (57) ms; p = 0.05] and lateral [1380 (71) ms vs 1322 (68) ms; p = 0.04] myocardial native T₁ compared to LGE− boys with DMD respectively (Table 4). No significant septal myocardial nor within-slice SD differences were observed between LGE− and LGE+ boys with DMD (Table 5). Lateral myocardial native T₁ was significantly increased in LGE− patients compared to healthy controls [1322 (68) ms vs 1277 (58) ms; p = 0.02; Fig. 4]. Figure 4 also shows within-group regional T₁ differences for the three groups (controls, LGE+ DMD, and LGE− DMD). Compared to the lateral myocardium, the septal region had significantly lower pre-contrast T₁ values in both DMD subgroups. No significant within-group regional differences were observed in the healthy myocardium. LGE− boys with DMD with normal LVEF were compared against LGE− boys with reduced LVEF, but no significant differences were observed in native T₁ measurements [1315 (87) ms vs. 1308 (24) ms; p = 0.93], respectively.

Figure 5 displays regional post-contrast T₁ and ECV measurements from the cohort of boys with DMD. The following septal, and lateral myocardial post-contrast T₁ values were observed: 639 (112) ms, and 591 (128) ms, respectively. A pattern of decreased post-contrast T₁ in the lateral wall compared to measurements in the septal wall was observed, but this difference only reached significance in LGE+ boys with DMD [542 (93) ms vs. 613(134); p ≤ 0.05]. This pattern of shortened post-contrast T₁ in lateral myocardium is also clearly depicted in Fig. 3.

ECV measurements demonstrated significant differences between the septal and lateral myocardium in patients with DMD [27 (3) % vs. 30 (8) %; p = 0.001; Fig. 5]. This result is consistent with diffuse fibrosis and extracellular expansion occurring in the DMD disease process [4]. LGE + boys had a significantly increased lateral myocardial ECV [38 (7) % vs. 29 (6) %, p = 0.001; Table 4 compared to LGE− boys. However, at the septal level, no significance was reached for ECV [27 (4) % vs. 27 (4) %, p = 0.73; Table 4] comparison between LGE+ and LGE− boys with DMD. Similar to pre-contrast T₁, no significant differences were observed in post-contrast T₁ [641 (103) ms vs. 610 (178) ms; p = 0.49] or ECV [28 (4) % vs. 27 (3) %; p = 0.57] measurements between LGE− boys with normal LVEF and LGE− boys with reduced LVEF.

Significant correlations between T₁ mapping and metrics of LV function in DMD patients were observed. In the LGE+ group, pre-contrast T₁ and LVEDVI [R² = 0.68, p = 0.01], native T₁ and LVMI [R² = 0.56, p = 0.02], and post-contrast T₁ and LVEF [R2 = 0.53, p = 0.03] were significantly correlated. In LGE− boys with DMD, a significant correlation was observed between post-contrast T₁ and LVEF only [R² = 0.49, p = 5.4 × 10^–3]. No T₁ and functional metrics were correlated in healthy controls. Additional file 2: Figure S2 A-D illustrates the significant correlations mentioned above.

The ROC evaluation revealed that a binomial logistic regression classifier using each biomarker in combination with the age, BMI, heart rate, and LVMI as features in all classification tasks resulted in a better model performance than each biomarker alone. Furthermore, the ROC analysis illustrated that all T₁-mapping biomarkers and LVEF are significant predictors of DMD and LGE status (AUC > 0.50); Fig. 6 displays the ROC curves for all the classification tasks. In the task of distinguishing between boys with DMD and healthy controls, native T₁ was comparable to LVEF (AUC = 0.88 vs. AUC = 0.87), but the combination of pre-contrast T₁ and LVEF yielded the best performance (AUC = 0.93). Figure 6b displays the same behavior in the LGE− vs. LGE + boys with DMD using native T₁, LVEF, and the combination of the two biomarkers (AUC = 0.84 vs. AUC = 0.83 vs. AUC = 0.87), respectively. In the task of predicting LGE status, ECV (AUC = 0.95) outperformed pre- and post-contrast T₁ (AUC = 0.83, AUC = 0.93), and LVEF (AUC = 0.84). The combination of native T₁ and LVEF (AUC = 0.88) again, performed better in the task of distinguishing between LGE + vs. LGE− boys with DMD compared to each biomarker performing individually.