Myocardial T1-mapping at 3T using saturation-recovery: reference values, precision and comparison with MOLLI

Background Myocardial T1-mapping recently emerged as a promising quantitative method for non-invasive tissue characterization in numerous cardiomyopathies. Commonly performed with an inversion-recovery (IR) magnetization preparation at 1.5T, the application at 3T has gained due to increased quantification precision. Alternatively, saturation-recovery (SR) T1-mapping has recently been introduced at 1.5T for improved accuracy. Thus, the purpose of this study is to investigate the robustness and precision of SR T1-mapping at 3T and to establish accurate reference values for native T1-times and extracellular volume fraction (ECV) of healthy myocardium. Methods Balanced Steady-State Free-Precession (bSSFP) Saturation-Pulse Prepared Heart-rate independent Inversion-REcovery (SAPPHIRE) and Saturation-recovery Single-SHot Acquisition (SASHA) T1-mapping were compared with the Modified Look-Locker inversion recovery (MOLLI) sequence at 3T. Accuracy and precision were studied in phantom. Native and post-contrast T1-times and regional ECV were determined in 20 healthy subjects (10 men, 27 ± 5 years). Subjective image quality, susceptibility artifact rating, in-vivo precision and reproducibility were analyzed. Results SR T1-mapping showed <4 % deviation from the spin-echo reference in phantom in the range of T1 = 100–2300 ms. The average quality and artifact scores of the T1-mapping methods were: MOLLI:3.4/3.6, SAPPHIRE:3.1/3.4, SASHA:2.9/3.2; (1: poor - 4: excellent/1: strong - 4: none). SAPPHIRE and SASHA yielded significantly higher T1-times (SAPPHIRE: 1578 ± 42 ms, SASHA: 1523 ± 46 ms), in-vivo T1-time variation (SAPPHIRE: 60.1 ± 8.7 ms, SASHA: 70.0 ± 9.3 ms) and lower ECV-values (SAPPHIRE: 0.20 ± 0.02, SASHA: 0.21 ± 0.03) compared with MOLLI (T1: 1181 ± 47 ms, ECV: 0.26 ± 0.03, Precision: 53.7 ± 8.1 ms). No significant difference was found in the inter-subject variability of T1-times or ECV-values (T1: p = 0.90, ECV: p = 0.78), the observer agreement (inter: p > 0.19; intra: p > 0.09) or consistency (inter: p > 0.07; intra: p > 0.17) between the three methods. Conclusions Saturation-recovery T1-mapping at 3T yields higher accuracy, comparable inter-subject, inter- and intra-observer variability and less than 30 % precision-loss compared to MOLLI. Electronic supplementary material The online version of this article (doi:10.1186/s12968-016-0302-x) contains supplementary material, which is available to authorized users.


Background
The recent introduction of rapid parameter mapping into cardiovascular magnetic resonance (CMR) imaging provides the invaluable ability for noninvasive quantitative myocardial tissue characterization. The quantification of the native longitudinal magnetization recovery time as a spatially resolved map (native T 1 -mapping) shows promising prognostic and diagnostic value in various cardiomyopathies [1]. The combination with post-contrast T 1 -time measurements allows for the estimation of the extracellular volume fraction (ECV), which reflects fibrotic remodeling [2], a common endpoint of many pathological cardiac conditions [3].
A number of cardiac T 1 -mapping methods have been proposed, each offering a distinct profile of advantages. The modified Look-Locker inversion recovery (MOLLI) sequence [4] and variations thereof, like the shortened MOLLI (ShMOLLI) [5], are commonly used for myocardial T 1 -mapping. However, confounding factors to the method's quantification accuracy including heart rate [6], T 2 relaxation time [7], and magnetization transfer [8] lead to underestimation of the T 1 -time of the healthy myocardium by~20 % at 1.5T [9,10].
Alternatively, saturation-recovery (SR) based myocardial T 1 -mapping methods have been proposed [11] and were recently revisited by the SAturation-recovery single-SHot Acquisition (SASHA) sequence [12]. To increase the low dynamic range in SR T 1 -mapping, the hybrid sequence for Saturation Pulse Prepared Heart-rate independent Inversion-REcovery (SAPPHIRE) T 1 -mapping was introduced, using a combination of saturation and inversion pulses for magnetization preparation [6]. While SASHA and SAPPHIRE result in excellent accuracy, the sequences still suffer from reduced precision in assessing T 1 -times compared with MOLLI, as previously shown at 1.5T [9].
The application of inversion-recovery T 1 -mapping at 3T has recently received increasing interest. Multiple studies have shown promising T 1 -map quality and improved quantification precision, due to the increased imaging Signal-to-Noise ratio (SNR) at 3T [5,13,14]. Thus, in this work we sought to study the visual quality and precision of SR T 1 -mapping and to establish accurate reference values for native T 1 -times and ECV-values of the healthy myocardium at 3T.

Methods
All images were acquired on a 3T MRI scanner (Magnetom Skyra; Siemens Healthcare, Erlangen, Germany) with a 30-channel receiver coil array.
Sequences T 1 -mapping was performed using the SAPPHIRE and SASHA SR methods and T 1 -times were compared to MOLLI T 1 -mapping. All T 1 -mapping sequences were implemented with a balanced Steady-State Free-Precession image acquisition (bSSFP) and shared the following parameters for phantom and in-vivo imaging: TR/TE/α = 2.6 ms/1.0 ms/35°, in-plane resolution = 1.7 × 1.7 mm 2 , slice-thickness = 6 mm, field-of-view = 440 × 375 mm 2 , bandwidth = 1085Hz/px, number of k-space lines = 139, linear profile ordering, startup-pulses = 5 Kaiser-Bessel, GRAPPA-factor = 2. The 5(3)3 MOLLI scheme was employed for native T 1 -mapping and the 4(1)3(1)2 scheme for post-contrast imaging [15]. For SAPPHIRE and SASHA, 10 images were acquired with the 9 recovery times (inversion or saturation times, respectively) linearly spaced between the minimal (113 ms) and the maximum recovery time, as determined by the duration of the respective R-R interval. Magnetization saturation was achieved using a composite "Water suppression Enhanced through T 1 -effects" (WET) [16] saturation module. An adiabatic full passage tan/tanh pulse [17] was used for magnetization inversion.

Phantom experiments
Phantom scans were performed to study pulse-efficacy, ex-vivo accuracy and precision of the SR T 1 -mapping methods at 3T. Detailed description of the phantom experiments can be found in the Additional file 1.

In-vivo experiments
20 healthy volunteers (27 ± 5 years, ranging from 20 to 39 years; 10 male: 27 ± 6 years; 10 female: 27 ± 4 years) were recruited for native and post-contrast T 1 -mapping. Figure 1 depicts the schematic of the scan protocol: A blood sample was drawn prior to each examination to measure blood hematocrit for ECV calculation and to exclude impaired renal function before the administration of a gadolinium based contrast agent (GBCA). Imaging was performed before bolus administration of 0.2 mmol/kg gadoterate meglumine (Dotarem; Guerbet, Aulnay-sous-Bois, France), and 15 and 25 min thereafter. T 1 -maps were acquired in three short-axis slices. Based on bSSFP frequency scout images, frequency offsets in the range of ±50 Hz and ±100 Hz were selected for MOLLI and the SR methods, respectively. Different off-resonance frequency shifts were chosen for MOLLI and SR, due to previously reported off-resonance sensitivity for the MOLLI sequence and off-resonance resilience for SR methods [10]. Post-contrast scan order was randomized to mitigate T 1 -trends caused by GBCA washout (Fig. 1).

Post-processing
Motion correction (MoCo, Advanced Retrospective Technique; Siemens Healthcare, Erlangen, Germany) was applied to co-register the T 1 -weighted image series. T 1 -maps were generated a) from the MoCo image series when the registration algorithm reduced residual motion and b) from the uncorrected image series when MoCo introduced registration distortions, as judged by visual assessment in consensus agreement of two reviewers (SW; 6 years of CMR experience, NMM; 4 years of CMR experience).
MOLLI T 1 -times were obtained using standard postprocessing [4] and SR T 1 -maps were generated using a three-parameter fit [18]. Regional ECV-values were calculated segment-wise according to the AHA 16segment model [19] for both post-contrast time points. Contrast agent concentrations were calculated for the myocardium and the blood-pool, based on the difference in the native and post-contrast relaxation rates (1/T 1 ) divided by an assumed relaxivity of 3.5 mmol/L/s [20].

T 1 -map analysis
Quantitative evaluation of T 1 -times and ECV-values was performed on a per-segment basis. In-vivo precision was defined as the intra-segment variation, measured in terms of standard deviation. Visual T 1 -map quality was evaluated by two readers, which were blinded to the sequence type (JB; >5 years of CMR experience; DL; >11 years of CMR experience). Each slice was scored separately with respect to overall T 1 -map quality (1: poor -4: excellent) [15] and visual off-resonance artifacts in the T 1 -map (1: strong artifacts -4: none). The detailed scoring criteria can be found in Additional file 2.
Average T 1 -times and ECV-values were statistically compared on a per-subject basis among the methods using ANOVA, followed by pair-wise paired Student's t-tests, if significant differences among the methods were detected. The inter-subject variability of the T 1 -times and ECV-values was compared among the methods using a Bartlett-test, and paired F-tests in case of significant results of the former. ECV-values between the two post-contrast time points were compared using a paired Student's t-test for each method. Furthermore, interand intra-observer variability was studied for native and post-contrast T 1 -mapping with the three sequences. A total of three ROI sets was independently drawn by two readers for each sequence and time point (Reader 1: UM, 12 years of CMR experience, Reader 2: NMM, 4 years of CMR experience; Reader 1: ROIs A, Reader 2: ROIs B, ROIs C). T 1 -times obtained with different ROI sets were compared on a per subject-basis for inter-(ROIs A vs. ROIs B) and intra-observer (ROIs B vs. ROIs C) analysis. Observer agreement was studied by analyzing the absolute difference between the T 1 -times as proposed in [21]. Observer consistency was assessed using the intraclass correlation coefficient (ICC) based on Winer's adjustment for anchor points [22]. The T 1 -time variation and the ordinal scaled image ratings were statistically evaluated using Kruskal-Wallis tests with subsequent Mann-Whitney U tests in case of significant difference between the three methods. Differences in the observer agreement were assessed with one-way analysis of variance (ANOVA) of the log-transformed absolute difference [22]. ICCs were statistically compared using two-tailed F-statistics, with Bonferroni correction yielding significance for p < 0.017. All other statistical tests were performed at a significance level of p < 0.05.

Phantom experiments
WET saturation modules resulted in average saturation efficacy >99 % across a broad T 1 -range. The SR methods showed excellent accuracy (<3.9 % deviation). T 1 -time variation was 29 and 50 % lower using MOLLI compared with SAPPHIRE and SASHA, respectively.

In-vivo experiments
Scanning was successfully completed in all subjects, with no pathological findings. Eight (0.09 %) out of a total of 8640 segments were excluded from further analysis due to imaging artifacts (SAPPHIRE: 4, 0.14 %; SASHA: 4, 0.14 %). Post-contrast results are given for the first time point (~15 min) in the remainder of the study if not explicitly stated otherwise. Fig. 1 In-vivo imaging protocol: After blood draw, all subjects underwent MR examination of approximately 1 h duration, including T 1 -mapping sessions prior to, 15 and 25 min after GBCA injection. Basic adjustments and frequency scouting were performed before native T 1 -mapping. To minimize the effects of GBCA washout on inter-sequence comparison, measurements of the same slices were grouped. The sequence orders within the group, as well as the slice order were randomized for each subject Figure 2 shows exemplary native and post-contrast T 1maps acquired with MOLLI, SAPPHIRE and SASHA in two healthy subjects. All three methods depict a homogeneous myocardium clear of artifacts.
The MOLLI in-vivo variation (53.7 ± 8.1 ms) shows no significant difference (p = 0.057) compared with SAPPHIRE (60.1 ± 8.7 ms), but a significant reduction compared with SASHA (70.0 ± 9.3 ms, p < <0.001). SAPPHIRE yields lower variation than SASHA (14 ± 10 %, p < 0.002). Both SR T 1mapping methods show a trend of increased variation in the inferior, inferior-lateral and anterior segments, compared with the septal and anterior-lateral segments. Figure 3 (bottom row) shows the segmental ECV based on the first post-contrast session. MOLLI yields the lowest ECV-values, followed by SAPPHIRE and SASHA, with all differences being significant (p < 0.007). There was no significant difference between the inter-subject variability of the ECV-values obtained with MOLLI (0.026), SAPPHIRE (0.020) and SASHA (0.025) (p = 0.53).
A summary of T 1 -times and ECV-values for the native myocardium and both post-contrast times are given in Table 1. The second post-contrast imaging time showed a trend of higher ECV-values than the first time point, with an absolute deviation of 0.014 ± 0.016 for MOLLI (p < 0.001), 0.008 ± 0.013 for SAPPHIRE (p < 0.02), and 0.005 ± 0.020 for SASHA (p = 0.24).
The results from the inter-and intra-observer analysis are given in Table 2. High reproducibility in terms of agreement was shown for all three sequences, with mean differences <15 ms for native and <6 ms for post-contrast T 1 -mapping. Good intra-observer consistency was obtained in all scans (ICC > 0.97). Inter-observer consistency was slightly lower across the three sequences, especially for native T 1 -times (ICC > 0.94). No statistically significant difference was found among the three sequences, neither in terms of agreement nor consistency. Figure 4 depicts the readers' quality and artifact scores. All three sequences were scored with "good" image quality on the average. MOLLI resulted in the highest average quality scores, followed by SAPPHIRE. The SASHA method showed the lowest quality in visual assessment. All pair-wise differences were found to be significant (p < 0.03). The average artifact scoring was significantly better for MOLLI (3.6 ± 0.3) compared with SAPPHIRE (3.4 ± 0.3, p = 0.03) and SASHA (3.2 ± 0.3, p = 0.001). Example images illustrating the effect of off-resonance artifacts on the T 1 -maps are given in Additional file 3: Figure S3.
MoCo was successfully performed on almost all MOLLI imaging series (97 %) and on the majority of the SAPPHIRE data (82 %). However, only few SASHA imaging series were correctly registered using MoCo (8 %).

Discussion
In this study, we assessed reference values and in-vivo precision of SR T 1 -mapping at 3T in comparison with MOLLI. SR T 1 -mapping provided robust image quality throughout the study. MOLLI T 1 -maps were shown to consistently provide the highest image quality rating and lowest artifact incidence. However, significantly better ex-vivo accuracy was confirmed for SR methods for the trade-off against a slight reduction of in-vivo precision. No significant difference was found in the inter-subject variability and the inter-and intra-observer variability among the three methods.   Fig. 3 Bullseye plots comparing the native T 1 -times (top row), precision (middle row) and the ECV-values (bottom row) of the three T 1 -mapping sequences averaged over all volunteers. The given ECV-values were calculated from post-contrast T 1 acquired 15 min after GBCA injection. Segmentation was performed according to the AHA 16-segment model in three short-axis slices (A = apical, M = mid-ventricular, B = basal). The average across all segments is given in the center of the bullseye, the slice averages can be found below. The MOLLI sequence shows lower T 1 , better precision and higher ECV-values compared to the saturation-recovery methods. SAPPHIRE results show similar native T 1 -and ECV-values with slightly better precision compared with SASHA Native T 1 -times of the human myocardium using SR T 1 -mapping were found to be around 1550 ms. This reveals a field strength dispersion of approximately 30 % compared with 1.5T (1210-1220 ms [9]), which is in good agreement with reported literature values for cardiac tissue of animals [23,24]. MOLLI T 1 -times from our own findings and previous reports at 3T (1166 ms [14]) demonstrate a significant underestimation of about 20-30 % compared with the present results of SR T 1mapping. This underestimation, as confirmed by the phantom study, indicates decreased in-vivo accuracy of MOLLI. SASHA T 1 -mapping was previously reported to have about 150 % higher in-vivo variability than MOLLI at 1.5T [10]. Our results demonstrate that the loss in precision when using SR over MOLLI is drastically reduced compared with 1.5T. The present results indicate that at 3T, MOLLI remains to provide higher visual image quality than SR methods. However, the high exvivo accuracy, the low level of precision-loss, and the good inter-subject variability, indicate only a small gap to SR T 1 -mapping. Hence, SR methods at 3T provide a valuable option for trading-off increased quantification accuracy against a reduction of overall image-quality.
Alternative T 1 -map reconstructions have been proposed for SR T 1 -mapping, to improve precision albeit at the cost of reduced accuracy and increased sensitivity of  Fig. 4 Pie charts showing the distribution and average of the quality (top row) and artifact (bottom row) scoring across all T 1 -maps and across both readers. Eighty-one percent of all images were scored with at least "good" image quality, with MOLLI having the highest average score and the lowest artifact scoring. SAPPHIRE shows higher average quality and similar artifact scores compared with SASHA the T 1 -time to the choice of scan parameters. A two-parameter fit for SASHA T 1 -mapping was recently proposed [10] and initial results on an extension using a variable flip-angle scheme for the bSSFP imaging readout to minimize the loss in accuracy were presented [25]. Two parameter fitting has also been used for SAP-PHIRE post-contrast T 1 -mapping [6]. However, imperfect inversion efficiency might impair the accuracy of SAPPHIRE, when using a two-parameter fit for native T 1 -mapping. The use of a predetermined correction factor for incomplete inversion, as previously proposed [17], might be warranted for this application. The reported reference ECV values for MOLLI (~0.26) and SR T 1 -mapping (~0.21) obtained in this study are in good agreement with previous literature. For MOLLI, ECV-values between 0.25 and 0.27 have been reported at 1.5T [9,15,26] and between 0.26 and 0.28 at 3T [26][27][28][29]. Furthermore, the slight increase in ECV-values between the two post-contrast time points has been previously observed with MOLLI at 3T [29], and the ECV deviation between the two time points is in agreement with previous reports (0.258-0.272 for times between 10 and 25 min [28]. Close agreement of SAPPHIRE ECV-values are obtained with a previous study at 1.5T (ECV: 0.20 [9]). SASHA ECV-values were reported as 0.18 [9], 0.21 [30] and 0.22 [31,32] in healthy subjects at 1.5T. The close agreement of these values with our results, as well as with ECV-values obtained with SR T 1 -mapping in an animal study at 3T (AIR: 0.20-0.21 [33]) proves high cross fieldstrength consistency for SR based ECV-measures.
Despite the higher precision of MOLLI compared to SR T 1 -mapping, previous studies did not report significant differences in the scan-rescan reproducibility [9,26]. To add on this, our results show no significant difference in the inter-or intra-observer variability between the methods either. All three methods showed consistency with ICCs > 0.90, which is considered excellent for diagnostic tools [34]. The values of observer variability characteristics obtained in this study are well in line with previous reports [27,[35][36][37][38]. However, some studies from specialized centers achieved consistently higher inter-and intra-observer variability, with ICCs > 0.99 [27,29,39]. This difference might be explained by the limited clinical experience of our readers. Therefore, extensive observer training and an extensive common learning phase for both readers seems to be required to achieve optimal reproducibility results in T 1 -mapping.
Imaging at 3T using bSSFP has considerable challenges compared with 1.5T. Off-resonance artifacts are commonly induced by magnetic susceptibilities at tissue interfaces, e.g. epicardium-lung interface. In this study, frequency scouts were used to minimize off-resonance artifacts. However, careful volumetric shimming is still essential at 3T to ensure robust image quality. Also, the rapid imaging readout reaches specific absorption rate (SAR) limitations at 3T. As SR T 1 -mapping methods were shown to be independent of the imaging flip-angle [12], improved imaging SNR could potentially be achieved using optimized excitation pulses with higher flip-angles and low SAR, for the trade-off against suboptimal slice profiles.
Non-rigid motion correction algorithms, as used in this study, are dependent on strong contrast within the area of interest [40]. Hence, MoCo was more effective for MOLLI than for the SR methods. Tailored motion correction algorithms might be required if a further reduction of residual motion in the SR imaging series is necessary.
This study has several limitations. Due to the lack of feasible methods for the assessment of "true" T 1 -times in the myocardium, no direct evidence of the in-vivo accuracy of SR methods can be given. Instead, phantom accuracy was used as an indicator of in-vivo accuracy. Evaluation of the sequence characteristics was restricted to accuracy and precision, specifically no inter-or intrasession reproducibility was considered in this study. A tightly controlled cohort of young healthy volunteers was recruited for the study, in order to obtain reproducible reference values of the healthy myocardium that are not affected by potential age-related fibrosis in the muscle. As the T 1 -time of the myocardium is known to be age and sex dependent [41], cohorts that are age/sex matched to the particular patient population are to be assessed if more specific T 1 -reference values with reduced intra-cohort variability are required.

Conclusions
Saturation-recovery at 3T was shown to provide accurate and robust T 1 -map quality at a field-strength of 3T. Invivo comparison to MOLLI showed decreased subjective image quality scores, a slight loss in precision, but comparable inter-subject, inter-, and intra-observer variability. Authors' contributions SW performed sequence implementation and first drafting of the manuscript and was engaged in conception and design of the study, data acquisition in phantoms and volunteers, as well as the analysis, interpretation and statistical analysis of data. NMM was responsible for study organization, volunteer recruiting and clinical concerns and was engaged in conception and design of the study, data acquisition in phantoms and volunteers, analysis, statistical analysis and interpretation of data, and revising and finalizing the manuscript. JB participated in study conception, medical and clinical concerns during study realization, data interpretation, and manuscript revision and evaluated quality and artifacts of all images. DL participated in study conception, evaluated quality and artifacts of all images and revised the manuscript for medical content. UM made substantial contributions to the analysis of the data and critically revised the manuscript. TP participated in study conception and was as clinical investigator responsible for medical and clinical concerns during study realization, data interpretation, and manuscript revision. FGZ was responsible for study ethics and organizational matters and critically revised the manuscript. LRS contributed by overseeing the study and editing various drafts of the manuscript. All authors read and approved the final manuscript.

Additional files
Competing interests SW is inventor of a pending U.S. and European patent entitled "Methods for scar imaging in patients with arrhythmia", which described the SAPPHIRE imaging sequence. No other financial or non-financial competing interests exist for any author.

Consent for publication
Not applicable.
Ethics approval and consent to participate This prospective study was approved by the local Institutional Review Board (Medizinische Ethikkommission II of the Medical Faculty Mannheim, Heidelberg University) and written informed consent was obtained from each subject prior to the imaging session.