Reproducibility of 3.0 T single-frequency cardiac MR elastography for quantification of left ventricular myocardial stiffness

Figures

• 

  Figure 1. The top row (A-E) shows a 26-year-old healthy male volunteer; the middle row (F-J) shows a 52-year-old man with aortic stenosis; and the bottom row (K-O) shows a 72-year-old female with long-standing hypertension. A, F, and K are end-diastolic short-axis cine images at the mid-left ventricular level. B, G, and L are magnitude images used for manual myocardial ROI delineation. C, H, and M show octahedral shear strain signal-to-noise ratio (OSS-SNR) maps within the propagated ROI. D, I, and N are gray-scale elastograms, and E, J, and O are color elastograms (color scale, 1.0-8.0 kPa). Mean left ventricular myocardial stiffness and OSS-SNR were extracted from the ROI. Myocardial stiffness and OSS-SNR were 7.6 kPa and 1.51 in the healthy volunteer, 11.1 kPa and 1.32 in the patient with aortic stenosis, and 11.1 kPa and 1.26 in the patient with hypertensive heart disease, respectively. These examples demonstrate increased myocardial stiffness in pressure-overload cardiomyopathy.

• 

  Figure 2. Inter-observer agreement between Reader 1 (R1) and Reader 2 (R2) at baseline scatterplots R1 measurements on the x-axis and R2 measurements on the y-axis (n = 69). Both metrics demonstrated excellent agreement: stiffness ICC = 0.970 [0.953, 0.982], r = 0.972; OSS-SNR ICC = 0.944 [0.911, 0.965], r = 0.969. Diagonal solid line = line of perfect identity (R1 = R2). Colored lines = linear regression with 95% CI (shaded); color encodes timepoint. Points colored by group (blue = healthy, pink = disease). ICC = intraclass correlation coefficient (2-way mixed, absolute agreement, single measures) with 95% CI. r = Pearson correlation; ***P < .001.

• 

  Figure 3. Bland-Altman plots for inter-observer agreement between Reader 1 (R1) and Reader 2 (R2) at baseline. (A) LV myocardial stiffness (kPa); (B) OSS-SNR. Bland-Altman plots show the difference (R1-R2) against the mean of both readers’ measurements (n = 69). The solid horizontal line denotes the mean bias; dashed lines indicate the 95% limits of agreement (LoA; bias ± 1.96 SD). Regression line with 95% confidence band (shaded) tests for proportional bias. Bias was negligible for both stiffness (+0.005 kPa) and OSS-SNR (−0.007). Statistically significant proportional bias was detected for both metrics (stiffness: r = −0.26, P = .033; OSS-SNR: r = −0.67, P < .001), although all differences remained within the 95% LoA. Points are colored by group.

• 

  Figure 4. Bland-Altman plots for intra-observer reproducibility of LV myocardial stiffness measurements. (A) Baseline vs. short-interval rescan (SIR, ∼5 minutes); (B) baseline vs. long-interval rescan (LIR, ∼1 month). Within each panel, Reader 1 (R1) and Reader 2 (R2) are displayed side by side. The solid horizontal line denotes the mean bias; dashed lines indicate the 95% limits of agreement (LoA). The regression line with 95% confidence band (shaded) tests for proportional bias. Bias values were negligible across all comparisons (range: −0.024 to +0.035 kPa). No statistically significant proportional bias was detected in any panel (all P > .77).

• 

  Figure 5. Absolute stiffness and OSS-SNR measurement errors between short-interval (SIR) and long-interval (LIR) rescans. (A) Absolute stiffness error (|Baseline – repeat scan|, kPa); (B) absolute OSS-SNR error (×0.001). Violin plots compare absolute errors between short-interval rescan (SIR) and long-interval rescan (LIR) for LV myocardial stiffness and OSS-SNR. Embedded box plots show the median and interquartile range (IQR); individual data points are overlaid on each distribution. For stiffness (A), Reader 1 demonstrated a statistically significant increase in absolute error from SIR to LIR (P = .003), whereas Reader 2 showed no significant difference (P = .948). For OSS-SNR (B), no significant SIR-to-LIR change was observed for either reader (R1: P = .178; R2: P = .436), confirming stable acquisition quality across rescan intervals.