Comparison of native myocardial T1 and T2 mapping at 1.5T and 3T in healthy volunteers

Cardiac magnetic resonance imaging (CMRI) is the gold standard in noninvasive myocardial tissue characterization [1]; however, early, subtle or diffuse myocardial tissue changes may not be evident using late gadolinium enhancement or conventional scanning techniques which depend on relative signal intensity differences between affected and adjacent, unaffected myocardial tissue [2]. Meanwhile, the use of relaxometry methods has gained widespread acceptance [2‐4]. Mapping techniques overcome the dependency on visually distinguishing differences in signal intensities. They allow direct measurement of T1 spin-lattice relaxation time or T2 spin-spin relaxation time in milliseconds and quantify the signal of each voxel on a standardized scale. Relaxation times depend on the water content in the tissue and allow further tissue characterization. Relaxation times vary in different tissues but also within the same tissue undergoing pathophysiological changes (e.g. inflammation, ischemia, edema and fibrosis) [2]. In addition, unenhanced T1 measurements enable the detection of various T1-altering substances, such as lipids in Anderson-Fabry disease, proteins in cardiac amyloidosis or iron deposits in hemochromatosis [5‐7]. Furthermore, it was shown that native T1-mapping measurements are significantly predictive of all-cause mortality and heart failure events in nonischemic cardiomyopathy [8].

The T2-weighted (T2w) CMRI and T2-weighted short tau inversion recovery (STIR) techniques are well-established sequences for the detection of myocardial edema [9] to differentiate acute from chronic myocardial changes [10, 11]. The quantitative approach of T2-mapping is a robust, accurate and fast alternative to conventional T2w and STIR sequences for the detection of edema in acute inflammatory cardiomyopathies. This method has the potential to even monitor myocardial inflammation in healing myocarditis [12], which has important clinical implications. Furthermore, it plays a role in the diagnosis of myocardial infarction by overcoming some restrictions of conventional T2w imaging [4], such as sensitivity to myocardial motion, surface coil intensity variation, high subendocardial signal from static blood, incomplete blood suppression and inter-reader or intra-reader variability [13].

An important aspect of mapping techniques is that the obtained information has prognostic relevance. The pathology cannot only be detected but also be quantified and the course of disease can be monitored under treatment. Therefore, mapping techniques have the potential to act as biomarkers to facilitate diagnostic decision making (e. g. iron overload, Anderson-Fabry disease) [14].

Despite promising results in the literature, routine implementation of this technique remains challenging. This might be associated with conflicting data regarding normal values. Recent publications suggest that even in sophisticated facilities, a variety of technical, image acquisition and patient-dependent factors influence T1 and T2 relaxation times [15]. To evaluate this issue the implementation of relaxometry methods in a general radiology department performing approximately 500 CMRI annually was investigated. To obtain reliable standard values for both 1.5T and 3T, the T1 and T2 relaxation times were measured in the same cohort using established state of the art mapping sequences.

All 120 CMRI scans were carried out without adverse events. The scan was incomplete in one subject due to technical problems and two subjects had to be excluded due to the detection of a chronic disease (sarcoidosis and an atrial septal defect II, respectively). Therefore, 115 examinations in 58 subjects were available for further analysis. The T1 and T2 maps at 1.5T were available for 57 subjects, T1 and T2 maps at 3T were available for 58 subjects. Due to malpositioning of the basal slice too close to the atrioventricular plane 15 segments (1.6%) had to be eliminated from the T1 measurements at 1.5T, 10 segments (1.1%) from T1 at 3T, 15 segments (1.6%) from T2 at 1.5T and 21 segments (2.3%) from T2 at 3T measurements. Females and males showed no statistically significant differences with respect to age, heart rate and body mass index (Table 1).

This study was designed to establish reference values for T1 and T2 relaxation times in healthy volunteers for our specific setting at 1.5T and 3T. Furthermore, the correlation between the intra-individual global T1 and T2 values at different field strengths was studied and published data concerning possible influences of physiological factors on relaxation times were reproduced.

As already pointed out in previous publications, different mapping techniques and field strengths are associated with a variety of normal values [20]. These values were comparable with publications using similar mapping protocols showing an overlap in confidence intervals; however, differences were also observed even in comparison with studies using MOLLI techniques for T1 and GraSE techniques for T2 mapping which were applied in this cohort ([20]; Table 2). As indicated by Kellman and Hansen [23] reasons for this inconsistency might be due to small variations in the study protocols (e. g. different flip angle, differences in echo-spacing leading to partial volume errors, shimming, different MOLLI scheme). This has an important implication for clinical practice as apparently minor changes in mapping protocols might have an extensive impact on relaxation times. A recent study of Roy et al. [24] providing normal values using a 3T scanner from the same manufacturer but with a different MOLLI scheme with a 3(3)3(3)5 bSSFP sequence showed considerably lower T1 relaxation times of the whole heart than this study using a 5(3)3 bSSFP scheme (1122 ± 57 ms vs. 1184 ± 38 ms). This observation underlines that reference values have to be reassessed individually in every radiological department before implementing mapping techniques for routine CMRI examinations. The authors propose that a calibration of mapping protocols is necessary if myocardial T1 and T2 relaxation times are to be used as MRI biomarkers in clinical practice.

Apart from differences in study protocols patient-related physiological factors are also presumed to affect relaxation times. Of note, various studies found an association between aging and lower T1 and higher T2 relaxation times [17, 21, 25]. In this institute more than two thirds of the CMRI clientele are below the age of 60 years. To avoid age-related alterations, this study focused on a younger population (only 4 subjects above the age of 60 years with a mean age of 41 years, ranging from 20 to 69 years). Consistently, in this cohort no significant differences of T1 and T2 values between age groups (<45 years of age vs. ≥45 years of age) were found. Nevertheless, besides age gender and heart rate have also been discussed controversially as possible influencing factors for relaxation times [13, 21, 25, 26].

In this cohort, gender was a confounding factor for T1 and T2 relaxation times at 1.5T and 3T. According to previous studies, females showed a significant prolongation of T1 relaxation times at 1.5T and 3T and of T2 relaxation times at 1.5T [17, 26‐28]; however, in accordance with von Knobelsdorff-Brenkenhoff et al. and Roy et al. [21, 24] no significant prolongation of T2 relaxation times in females at 3T was observed which might be associated with numerous artifacts at 3T in this study (Fig. 3). Furthermore, a heart rate ≥70 bpm led to a significant global prolongation of T1 relaxation times at both field strengths. Even though the MOLLI 5(3)3 schema was used which is supposed to be less sensitive to higher heart rates plus a frequency adapted algorithm for higher heart rates, an effect seems to remain [29]. A possible explanation might be the influence of partial volume effects due to relatively thick slices (8 mm) and cardiac motion artifacts especially in thin myocardial walls [23].

The prolongation of T1 relaxation times in higher heart rates is in accordance with Piechnik et al. [26] while others described no significant influences [21]. In accordance with previous studies a heart rate of ≥70 bpm was associated with a significant reduction of T2 relaxation times at 3T [21]. The small but significant influence of heart rate on T1 relaxation times and on T2 relaxation time at 3T leads to the assumption that an algorithm for compensation of higher heart rates is recommended to improve mapping quality for detection of subtle myocardial changes.

While a correlation of relaxation times at different field strengths in the same individual is presumed, to the best of our knowledge no study thus far has investigated this issue in a larger cohort. Therefore, we focused on this interesting subject and indeed were able to verify this association. This means that if a subject had low mean T1 values at 1.5T, these values were also low at 3T. As there are limited data on the reproducibility of myocardial T1 and T2 relaxation times at different field strengths in the same individual this is an important finding. The correlation was higher for T1 than for T2 values (Fig. 9). This might be associated with artifacts affecting T2 measurements especially at 3T, primarily off-resonance artifacts in regions adjacent to the lungs in addition to other artifacts, such as partial volume and motion artifacts. Of note, this is a critical point that becomes more relevant when evaluating the same individual at different field strengths. According to other publications, myocardial T1 relaxation times vary significantly with field strength. An increase of T1 values and a decrease of T2 values with higher field strength is reflected in previous studies [20, 28, 30‐34]. The mean T2 value at 1.5T using a GraSE technique is consistent with other publications [4, 13, 22, 24, 35]; however, while at 3T the T2 relaxation time was comparable with Baessler et al. using a GraSE sequence as applied in this study, other authors described considerably lower T2 values using accelerated multi-echo spin echo (MESE) or T2-prepared bSSFP sequences [21, 36‐38]. Again, this finding highlights the necessity of individual evaluation of normal values in every radiological institution. Furthermore, one can speculate that if institutional reference values are not available, only critical comparison with published relaxation times using similar protocols may be considered.

As already noted by other authors, also in zjis cohort the intrasubject and intersubject variability of T1 and especially of T2 values was considerably high per slice and per segment ([4, 21, 38], Table 4; Figs. 5, 6 and 7). This suggests a broad spectrum of normal values probably due to different effects of artifacts (e.g. B0, B1 inhomogeneities, off-resonance artifacts, motion artifacts, susceptibility artifacts or partial volume) in different areas of the myocardium. Apart from artifacts it has to be considered that e. g. regional differences in extracellular volume (ECV), amount of collagen, collagen fiber orientation and regional perfusion influence normal values. This insight complicates the distinction between borderline and pathologic relaxation times.

Due to respiratory or cardiac motion artifacts and susceptibility artifacts, a proportion of segments had to be excluded from this analysis (Fig. 3). Of note, the T2 maps at 3T emerged as the most vulnerable sequence. In contrast to Baessler et al. fewer visual artifacts were obtained using a comparable GraSE sequence at 1.5T but exclusion rate at 3T was almost the same (6.4% vs. 20–24% at 1.5T and 14.4% vs. 13–15% at 3T) [38]. The main cause of artifacts are B0 and B1 inhomogeneities which are especially problematic at 3T leading to differences in the signal across images and between segmental measurements. Especially while using thick slice imaging partial volume effects cause artifacts particularly in apical slices and in hypermobile regions, such as the free lateral wall. Interestingly, as previously described [38], also in this study the elimination of segments affected by artifacts did not lead to exclusion of extreme value outliers.

In this study the topics of extracellular volume (ECV) and post-contrast T1-mapping were not elaborated, as it was pointed out in previous studies that native T1-mapping could differentiate between healthy and diseased myocardium with high sensitivity, specificity and diagnostic accuracy [8, 39]. Even though noninvasive quantification of myocardial ECV via post-contrast T1-mapping is well validated [40, 41], native T1-mapping represents an excellent, rapid, noncontrast alternative for the detection of myocardial tissue remodelling [42]. Furthermore, The MOLLI 5(3)3 scheme used in this institution is known to have an excellent precision especially for native T1-mapping but is not optimal for shorter T1 values associated with contrast [23].

There are several limitations to this study. This was a single center study of moderate sample size using MRI scanners from only one manufacturer. Therefore, a center-specific and manufacturer-specific bias cannot be excluded. No histological examination of the myocardium was performed. In the post-processing phase no program for motion correction was available which led to higher exclusion-rates of myocardial segments. Despite the relatively small size of the cohort results of studies using similar mapping techniques could be reproduced. This underlines the validity of these reference values and implications. As specified above, two thirds of the patients undergoing CMRI evaluation were below the age of 60 years. Therefore, this cohort consisted of younger volunteers, which might obscure a pre-described age dependency of relaxation times in the aged myocardium. These normal values can only be applied and generalized in younger patients in institutions using similar mapping techniques to ours.

The results obtained during daily clinical practice in healthy volunteers indicated that T1 and T2 relaxation times showed gender and heart rate dependencies, even in T1 maps using a less heart rate sensitive MOLLI variant. Due to artifacts caution should be applied especially while performing T2 mapping at 3T and while measuring relaxation times in the inferolateral wall. As already pointed out in the consensus statement by the Society for Cardiovascular Magnetic Resonance (SCMR) endorsed by the European Association for Cardiovascular Imaging (EACVI) [14], we also highlight that reference values have to be reassessed in every radiological institution before implementing mapping protocols in daily clinical practice. Furthermore, to the best of our knowledge, the presumed correlation of intra-individual global T1 and T2 values measured at 1.5T and 3T could be demonstrated for the first time in a larger cohort.