This study examined myocardial T1 and T2 mapping techniques at 3 T in a large sample of healthy volunteers. The main findings are: i) T2 and T1 mapping achieve a high grade of diagnostic image quality, although susceptibility artifacts entailed the exclusion of a limited number of myocardial segments from the analysis. ii) Observer dependency of T2 and T1 relaxation time quantification was low. iii) Mean values and 95% tolerance interval of myocardial T2 and T1 relaxation times are presented per slice and per segment and can be used as reference values specific for this MR setting.iii) An inter-subject distribution of T2 and T1 values became apparent and may constitute a limitation to define appropriate cut-offs.
Previous studies with SSFP-based T2 mapping at 1.5T did not report the exclusion of segments from analysis due to SSFP off-resonance or banding artifacts [2–4]. Hence, this challenge seems to surface at higher field strengths due to the increase in the peak-to-peak B0 inhomogeneity across the heart. The use of an appropriately selected delta frequency may be an option to resolve some artifacts and deserves further systematic investigation. The artifacts mainly affected the inferolateral region, where pathologies like myocarditis may also exhibit their predominant lesion . Despite that, the step from 1.5 T to 3 T for CMR is generally desired due to expected gains in signal, which may be exploited for improved spatial and temporal resolution. This potential promises to enable more detailed insights into cardiac tissue in order to facilitate the early detection of myocardial disease.
T2 relaxation times derived from T2-prepared SSFP imaging in this study are higher compared to a black-blood multi-echo spin-echo approach at 3 T, which provided a mean value of T2 = 39.6m sin the septum . Myocardial T2 reported here was found to be lower versus a mean T2 = 52.2 ms reported for T2 prepared SSFP imaging at 1.5 T . Possible explanations are: i) differences in the pulse sequence design, ii) differences in the spatial resolution, with lower resolution being associated with more partial volume and potentially higher T2 values, and iii) T1 relaxation effects due to higher T1 values at 3 T versus 1.5 T. Generally, myocardial T2 reported in the literature varies substantially, ranging from about 50 ms to 58 ms at 1.5 T . The heterogeneity of data underlines that the measured T2 relaxation time is very sensitive to cofactors and emphasizes the need to generate reference values specific for each technique and imaging setting.
Our results showed that T2 increased from base to apex, which is in accordance with a recent work using a similar mapping technique at 1.5 T . The most probable cause is partial-volume effects that increase towards the apex owing to the curvature of the left ventricle. To encounter this limitation, some groups exclude the apical slice from mapping to omit measurement errors . We tried to minimize this error by carefully drawing the contours in the middle of the myocardium while leaving out the endocardial portion of the myocardium, as well as by using an isotropic spatial resolution as high as possible.
Most of the previous studies reported T2 values averaged over all myocardial segments or only for a midventricular slice. By averaging T2 values over the whole slice or the whole heart, focal T2 deviations may be overlooked. The present study is the largest study, which reports T2 values for each myocardial segment and slice.
As reported for the global T2 values, the segmental T2 values increased from base to apex. In comparison, Markl et al. reported T2 values from 50.5 ms to 51.6 ms in the basal slice and 54.3 ms to 56.1 ms in the apical slice at 1.5 T .
The inter-subject variability of absolute T2 values was relatively large both per-slice and per-segment. This finding is in concordance with Thavendiranathan et al., who described T2 values ranging from about 50 ms to 62 ms in healthy controls , and with Giri et al., who reported that the apical region showed the most pronounced inter-subject variability . The high inter-subject variability can be considered as the main challenge of T2 mapping, given that the difference in T2 between healthy and injured myocardium has been reported to be relatively small, e.g. 13 ms/11 ms between infarct core/myocarditis and remote myocardium [3, 4].
The association of heart rate and T2 relaxation time is under discussion. Giri et al. reported that the variability between healthy subjects was unrelated to heart rate. Other studies reported lower T2 values in patients with higher heart rate [1, 4]. This may be attributed to the hypothesis that higher heart rates induce pronounced T1 relaxation effects caused by incomplete T1 relaxation, which may affect T2 mapping using a SSFP-based approach. This finding is very relevant for clinical practice as subtle T2 increases may disappear in acutely ill patients with higher heart rates.
T1 mapping demonstrated diagnostic image quality for the vast majority of myocardial segments. However, a relevant number of myocardial segments had to be excluded due to technical challenges, which would lead to diagnostic uncertainty in a clinical scenario. Previous studies at 1.5 T and 3 T reported lower rates of artifact-related non-diagnostic segments [7, 10, 16, 17]. The explicit source of the artifacts has not been reported in detail in most studies, which renders benchmarking against previous results challenging. A possible contributing factor might be that artifacts are often only visible in the original images - which are used for quality assessment - while they might be not apparent in the final maps. In our study, susceptibility artifacts in the inferolateral region were most frequent.
The pre-contrast T1 values are in concordance with Piechnik et al., who reported T1 = 1169 ms averaged over all myocardial segments . At 3.0 T higher midventricular T1 values (T1 = 1315 ms or T1 = 1286 ms) were reported when using a T1 mapping technique similar to that used in this study [17, 18]. These discrepancies underline that T1 relaxation times are sensitive to many influencing factors.
The myocardial T1 relaxation times reported here can be regarded as reference values specific only for this cohort, time point, mapping technique, type and dosage of contrast media. Further comparisons with other published results are difficult unless an identical study design is used. To provide a context, Lee et al. used 0.15 mmol Gadolinium DTPA and measured a mean T1 of about 550 ms in one midventricular slice after 8.5 min in healthy human subjects at 3 T .
We observed that the pre-contrast T1 times increased from base to apex, whereas the post-contrast T1 values decreased from base to apex. Partial-volume effects owing to the curvature of the left ventricle can most probably explain this finding with blood signal being included into the voxel. While some completely exclude apical T1 maps from analysis , we tried to minimize this error by excluding the endocardial portion of the myocardium and by choosing a high isotropic spatial resolution.
In agreement with Kawel et al. we did not observe significant segment-to-segment differences post-contrast . However, pre-contrast T1 values of the anterior segments were lower than T1 observed for the other segments. Interestingly, Piechnik et al. observed the identical pattern with MOLLI at 3 T . Kawel et al. confirmed the presence of regional variability of pre-contrast T1 values inspite of using a different classification into “septal” and “non-septal” myocardium . Although absolute regional difference was small, this finding has to be considered in clinical CMR interpretation as the difference between healthy and abnormal tissue might be in a similar range.
The inter-subject variability of absolute T1 values was notable both per-slice and per-segment, including extreme outliers. This finding is in concordance with other T1 mapping studies reporting pre-contrast T1 values at 1.5 T ranging from 862 ms to 1105 ms in healthy volunteers  and a coefficient of variation of 4.5% (pre-contrast) and 7.0% (post-contrast) . The high inter-subject range may be the main challenge of T1 mapping, given that the difference in T1 times between healthy and injured myocardium has been reported to be relatively small depending on the underlying disease. Dall’Armellina et al. reported a mean pre-contrast T1 value of 1257 ± 97 ms for acutely infarcted segments compared to 1196 ± 56 ms for normal unaffected segments at 3 T . In other myocardial diseases like Fabry’s disease or amyloidosis, pre-contrast T1 may already be accurate enough to differentiate cardiac amyloid patients from normals .
Post-contrast T1 in the present study was even more variable between subjects than pre-contrast T1, attributable to the many factors with influence on the contrast kinetics (e.g. patient weight, hematocrit, renal function). Miller et al. recently demonstrated that even though isolated post-contrast T1 measurement showed significant within-subject correlation with histological collagen volume fraction, the between-subject correlations were not significant. Hence, isolated post-contrast T1 measurement seems to be insufficient for assessing extracellular volume fraction .
Aging was found to be associated with decreasing pre-contrast T1 values. This is an interesting aspect that may reflect early age-dependent alterations of myocardial texture. Dall’Armellina et al. and Ugander et al. showed that pre-contrast T1 times were increased in acute myocardial ischemia [20, 23]. Dass et al. reported increase in pre-contrast T1 in cardiomyopathies. Hence, the present reduction of pre-contrast T1 with age may sound contradictory . In contrast, in a rat model, diffuse myocardial fibrosis was associated with a non-significant trend towards lower pre-contrast T1 values . Therefore our data are stimulating to further analyze the value of pre-contrast T1 mapping in non-ischemic heart disease in future.