Fig 1.
T1MES phantom used in this study, and the reproducibility study protocol.
a) The T1MES phantom used in this study consists of 9 vials of NiCl2 doped agarose covering T1 and T2 ranges in the blood and myocardium before and after Gadolinium-based contrast agents. b) An imaging session was repeated every 2–3 weeks over 20 months (between-session reproducibility). Within each session, slice-interleaved T1 and T2 mapping sequences were repeated 3 times (between-repetition repeatability) for five slices (between-slice repeatability). SE T1 and T2 measurements and MOLLI were performed for comparison. STONE-bSSFP, slice-interleaved T1 with balanced steady‐state free precession; STONE-GRE, slice-interleaved T1 with spoiled gradient echo; SE, spin-echo.
Fig 2.
Reproducibility between sessions, and repeatability between repetitions and slices of slice-interleaved T1 mapping sequences were assessed using coefficients of variation (CV).
Slice-interleaved T1 mapping sequences showed excellent between-session reproducibility (CV: SE T1 = 1.1±0.5%, MOLLI = 1.2±0.6%, STONE-bSSFP 2P = 0.8±0.4%, STONE-GRE 2P = 0.8±0.4%, STONE-bSSFP 3P = 1.0±0.3%, STONE-GRE 3P = 1.0±0.4%), between-repetition repeatability (CV: MOLLI = 0.5±0.6%, STONE-bSSFP 2P = 0.3±0.1%, STONE-GRE 2P = 0.3±0.1%, STONE-bSSFP 3P = 0.6±0.2%, STONE-GRE 3P = 0.6±0.2%), and between-slice repeatability (CV: STONE-bSSFP 2P = 0.3±0.2%, STONE-GRE 2P = 0.3±0.1%, STONE-bSSFP 3P = 0.6±0.3%, STONE-GRE 3P = 0.5±0.3%).
Fig 3.
Coefficients of variations (CV) shown as scatter plots for each vial.
Vials are sorted from shortest T1 to longest T1 time (reference T1 from the T1MES manual measured by slow inversion-recovery/spin-echo methods at 1.5T: 255, 300, 430, 458, 562, 803, 1090, 1333, and 1489 ms). Vials with higher T1 time show higher variability.
Table 1.
Sources of variability in T1 mapping defined by variance decomposition analysis.
Table 2.
Between-session reproducibility and the comparison of slice-interleaved T1 mapping sequences against SE T1 and MOLLI.
Table 3.
Linear regression analysis of slice-interleaved T1 mapping sequences against reference SE T1 measurements.
Fig 4.
Bland-Altman analyses of slice-interleaved T1 mapping sequences against SE measurements.
The mean difference (bias) is presented as the red line, and the 95% limits of agreement (mean ± 2 standard deviations) are presented as dashed lines. Each data point represents one study time point which was averaged for all repetitions and slices within each session. The T1 mapping sequences show underestimation compared to the reference measurement. STONE-GRE 3P shows strongest agreement with the reference measurement with an underestimation less than 1 ms.
Fig 5.
Reproducibility between sessions, and repeatability between repetitions and slices of slice-interleaved T2 mapping sequences were estimated using coefficients of variation (CV).
Slice-interleaved T2 mapping had good between-session reproducibility (CV: SE T2 = 2.5%, slice-interleaved T2 2P = 3.5%, slice-interleaved T2 3P = 6.3%, slice-interleaved T2 4-T2preps 2P = 4.7%, slice-interleaved T2 4-T2preps 3P = 6.7%), between-repetition repeatability (CV: slice-interleaved T2 2P = 2.7±0.2%, slice-interleaved T2 3P was 6.0±0.2%, slice-interleaved T2 T2preps 2P = 4.2±0.2%, slice-interleaved T2 T2preps 3P was 6.5±0.2%), and good between-slice repeatability (CV: slice-interleaved T2 2P = 2.9±0.2%, slice-interleaved T2 3P = 6.5±0.3%, slice-interleaved T2 T2preps 2P = 4.5±0.4%, slice-interleaved T2 T2preps 3P was 7.0±0.4%).
Table 4.
Sources of variability in T2 mapping defined by the variance decomposition analysis.
Table 5.
Linear regression of slice-interleaved T2 mapping sequences against the reference SE T2 measurements.
Fig 6.
Bland-Altman plots of slice-interleaved T2 mapping sequences against the reference SE T2 measurements.
Bland-Altman analyses of slice-interleaved T2 mapping shows an overestimation when the map is reconstructed with a 2-parameter fit model, and an underestimation when reconstructed with a 3-parameter fit model. Each data point represents one study time point which was averaged for all repetitions and slices within session. The mean difference (bias) is presented as the red line, and the 95% limits of agreement (mean ± 2 standard deviations) are presented as dashed lines.