CMR-based T1-mapping offers superior diagnostic value compared to longitudinal strain-based assessment of relative apical sparing in cardiac amyloidosis

Cardiac amyloidosis (CA) is an infiltrative disease. In the present study, we compared the diagnostic accuracy of cardiovascular magnetic resonance (CMR)-based T1-mapping and subsequent extracellular volume fraction (ECV) measurement and longitudinal strain analysis in the same patients with (a) biopsy-proven cardiac amyloidosis (CA) and (b) hypertrophic cardiomyopathy (HCM). N = 30 patients with CA, N = 20 patients with HCM and N = 15 healthy control patients without relevant cardiac disease underwent dedicated CMR studies. The CMR protocol included standard sequences for cine-imaging, native and post-contrast T1-mapping and late-gadolinium-enhancement. ECV measurements were based on pre- and post-contrast T1-mapping images. Feature-tracking analysis was used to calculate 3D left ventricular longitudinal strain (LV-LS) in basal, mid and apical short-axis cine-images and to assess the presence of relative apical sparing. Receiver-operating-characteristic analysis revealed an area-under-the-curve regarding the differentiation of CA from HCM of 0.984 for native T1-mapping (p < 0.001), of 0.985 for ECV (p < 0.001) and only 0.740 for the “apical-to-(basal + midventricular)”-ratio of LV-LS (p = 0.012). A multivariable logistical regression analysis showed that ECV was the only statistically significant predictor of CA when compared to the parameter LV-LS or to the parameter “apical-to-(basal + midventricular)” LV-RLS-ratio. Native T1-mapping and ECV measurement are both superior to longitudinal strain measurement (with assessment of relative apical sparing) regarding the appropriate diagnosis of CA.

Systemic amyloidoses comprise more than 30 different disorders characterized by an infiltrative deposition of misfolded protein in various organs such as skin, eyes, lung, liver, kidneys, nervous system and heart. Depending on the variety and degree of organ involvement, the respective clinical manifestation may differ 1 . Amyloidoses are classified based on the misfolded protein precursor-with light-chain (AL) amyloidosis and transthyretin (ATTR) amyloidosis being the most commonly found forms in case of cardiac amyloidosis (CA) 2 . Moreover, it is well known that cardiac involvement is associated with high morbidity and poor prognosis 3,4 . Hence, early and appropriate diagnosis of cardiac involvement is highly important in order to start a targeted therapy on time.
The accumulation of amyloid deposits in the extracellular space of the myocardium results in both increased biventricular wall thickness and ventricular stiffness 5,6 that are hallmarks of this restrictive cardiomyopathy, eventually leading to rapid progressive heart failure. Furthermore, deposition of amyloid fibrils in the atrial and ventricular wall may cause conduction abnormalities and contributes to the high prevalence of ventricular arrhythmias and atrial fibrillation, increasing the risk of sudden cardiac death (SCD) and thromboembolism 7 .
In contrast to CA, hypertrophic cardiomyopathy (HCM) is most often determined by mutations of those genes encoding sarcomere proteins of the contractile apparatus 8,9 . These mutations cause a disorganized arrangement of myocyte hypertrophy (called myocardial disarray) as well as expansion of extracellular matrix, composed of interstitial and replacement fibrosis 10 . Such structural changes in turn result in increased wall thickness and noncompliance of the left ventricle. Clinically, most patients with HCM are asymptomatic or show just mild symptoms, but a subset will progress to suffer from symptoms of heart failure, chest pain and arrythmias with an increased risk of SCD 11 . Cardiovascular magnetic resonance (CMR) has been established as an important diagnostic tool for the work-up of left ventricular hypertrophy (LVH) of unknown origin. In this context, different myocardial patterns of late gadolinium enhancement (LGE) allow to differentiate CA from HCM: While the LGE pattern in CA is rather diffuse, starts mostly from the subendocardial layer of the basal segments and eventually spreads to all myocardial layers and segments 12,13 , HCM is characterized by a patchy and more focally accentuated LGE pattern predominantly occurring in the most hypertrophic septal segments of the left ventricular myocardium 14,15 .
More recently, T1-mapping has emerged as a new diagnostic technique, which offers tissue characterization by measurement of the intrinsic T1 relaxation time of the myocardium 16 . T1-mapping before infusion of a gadolinium-based contrast agent (native T1) and thereafter (post-contrast T1) allows to determine the extracellular volume fraction (ECV) of the myocardium 17 . Both native T1 and ECV are useful tools for the work-up of hypertrophic cardiac phenotypes of unknown origin. Since CA is characterized by an extensive, diffuse amyloid infiltration of the extracellular space 18 while HCM mostly shows focally accentuated interstitial fibrosis in hypertrophied septal segments, the increase in global native T1 and ECV values is consistently lower in HCM as compared to CA 19 .
Another popular approach in the evaluation of infiltrative cardiomyopathies such as CA is based on tissue tracking techniques that aim to measure myocardial deformation kinetics and patterns (strain). Both echocardiography-and CMR-based strain techniques have been evaluated in different cardiac diseases. Particularly, speckle tracking echocardiography (STE) and CMR feature tracking (FT) are the most validated and clinically used methods for strain analysis 20 . In principle, in both diseases (CA and HCM) all strain parameters are somewhat reduced in advanced cardiac phenotypes. However, whereas the physiological gradient of baso-apically diminishing strain is mostly preserved in case of HCM, a distinctive and unique pattern is mostly observed in case of CA, i.e. an inversed pattern of gradual strain increase from the basal to apical segments of the left ventricle (LV), known as "cherry on top" or "apical sparing" phenomenon 21,22 .
With this study, we assessed the diagnostic value of T1-mapping-based approaches in comparison to strainbased techniques regarding the appropriate diagnosis and differentiation of CA and HCM.

Methods
Study population. This is a comparative, monocentric, prospective study. All patients underwent a routine CMR study for work-up of suspected non-ischemic cardiomyopathy. The first study group (CA group) comprised N = 30 patients with histologically proven cardiac amyloidosis (including both AL and ATTR subtypes). The second study group (HCM group) comprised N = 20 patients with "conventional" HCM showing preserved LV ejection fraction (LV-EF) ≥ 50%, LV wall thickness ≥ 15 mm (that could not be explained by abnormal loading conditions) and absence of LV outflow tract obstruction. The third group (control group) comprised N = 15 patients in whom structural or functional cardiac abnormalities were ruled out and who presented with a low pre-test probability of CAD. Exclusion criteria comprised: (1) relevant valvular disease (at least grade 2 in echocardiography and/or at least moderate in CMR), (2) prosthetic valve, (3) permanent atrial fibrillation and (4) obstructive coronary artery disease. The local ethics committee (Ethikkommission der Ärztekammer Westfalen-Lippe) approved the study protocol and every patient gave written informed consent before enrolment.

CMR acquisition.
All CMR studies were performed on a 1.5-T system (Ingenia, Philips Healthcare, Best, The Netherlands). CMR data acquisition was performed according to the standardized protocol suggested by the Society for Cardiovascular Magnetic Resonance (SCMR) 23 . Our CMR protocol comprised a cine steadystate free precession pulse sequence for ventricular function and a two-dimensional (2D) inversion recovery fast www.nature.com/scientificreports/ spoiled gradient-echo sequence 10 to 15 min after administration of gadolinium-based contrast (Gadobutrol 0.15 mmol/kg) for detection of myocardial pathology. Moreover, a modified Look-Locker inversion recovery (MOLLI) T1-mapping sequence was obtained in basal, mid and apical short-axes prior to contrast agent administration and ~ 20 min thereafter to determine native T1 and ECV values. Representative CMR acquisitions from all three groups of our cohort are illustrated in Fig. 1.
CMR data analysis. Image analysis and interpretation was performed using commercially available software (cvi42-version 5.11.0, Circle Cardiovascular Imaging, Calgary, Alberta, Canada). Analysis of ventricular volumes and function as well as LV mass was made by contouring short-axis cine images.
LGE images were visually assessed 24 and the "Query Amyloid Late Enhancement" (QALE) score was reported as described in more detail elsewhere 25 . www.nature.com/scientificreports/ T1 and ECV measurements. T1-mapping and ECV were assessed and reported based on the consensus statement of SCMR 26 . Motion corrected native and post-contrast T1 maps were generated from the pre-and post-contrast T1 sequences. In each short-axis T1 map the endo-and epicardial contours were manually drawn. In order to further hamper contamination from blood pool and neighbouring tissues, respectively, a 10% safety margin was automatically set for both contours. Further, each short-axis map was automatically segmented (6 segments) using the RV insertion points as reference. Additionally, for ECV calculation, a region of interest was drawn in the blood pool (avoiding the papillary muscles) in all analysed T1 maps. Motion corrected and segmented ECV maps were generated from the native and post-contrast segmented T1 maps, using the patient's haematocrit level.

Feature tracking analysis.
For the assessment of global LV deformation, three-dimensional (3D) LV global longitudinal strain (LV-GLS) derived from feature tracking (FT) was obtained using a validated algorithm integrated in the analysis software 27 . Landmarks for LV base (at the mitral valve ring) and apex were defined at end-diastole in all long-axis slices. Endocardial and epicardial borders were manually contoured in the end-diastolic frame in the three long-axis slices and in three short-axis slices, the most basal slice without through-plane distortion from the LV outflow tract, a mid-ventricular and an apical slice. Both the landmarks and the contours were automatically propagated throughout the cardiac cycle and manually corrected in case of inaccuracies. Subsequently, relative apical longitudinal strain (LS) was calculated based on the following equation: average apical LS/(average basal LS + mid LS), as defined by Phelan et al. 28 .
Statistical analysis. Statistical  Ethics approval and consent to participate. The study protocol complies with the Declaration of Helsinki. Written informed consent was obtained from every patient.

Results
Study population. The study group characteristics are summarized in Table 1. Males and females showed a similar distribution in the CA and HCM group (77% in the CA group vs. 80% in the HCM group; p = 0.13). Median age differed significantly between the CA and both the HCM and control group due to the higher prevalence of CA in elderly patients. There were no other significant differences in major cardiovascular disease risk factors that could theoretically influence the results of this study.
Conventional CMR findings. All anatomic, functional and structural CMR results are given in Table 2.   Table 3.

Discussion
The present study focuses on a head-to-head comparison of the diagnostic performance of T1-mapping-based vs. strain-based CMR parameters in CA and HCM patients. Our study results clearly show that the novel CMR parameters native T1 and ECV show a higher diagnostic yield in the delineation of CA from HCM patients (as well as controls) compared to a strain-based approach assessing the presence of "apical sparing". Today, FT-CMR is a popular technique for strain analysis that tracks the displacement of individual patterns representing anatomical structures over time 29 . A major advantage of this technique compared to other strain approaches is its ease-of-use as the analysis can be performed very quickly on standard cine-CMR images after the examination. Balanced steady-state free precession (b-SSFP) sequences that are currently routinely used for standard cine imaging provide excellent blood-myocardium contrast as well as a high spatial and temporal  www.nature.com/scientificreports/ resolution for motion tracking over the cardiac cycle. FT-CMR is most efficient in the detection of the endocardial border where the contrast between heart muscle and cavum is excellent. However, detection of the epicardial border can be challenging-in particular, when contrast between myocardium and surrounding tissue is poor. In some cases, tissue tracking may fail to find the transformation of every point within the myocardium because of the homogeneity of water content and properties of the tissue 30 . Moreover, it should be taken into account that a cine loop represents a reconstructed averaged cardiac cycle and thereby, in cases of high variation in heart frequency FT-CMR might give false measurements. In general, FT-CMR is a feasible and highly reproducible technique in various cardiac diseases. Noteworthy, assessment of LV-GLS by FT-CMR in case of LVH has shown good intra-und inter-observer variability with an intraclass correlation coefficient (ICC) of 0.82 (0.45-0.95) and 0.86 (0.54-0.96) respectively 31 . However, inter-study reproducibility seems to be poor (ICC: 0.44 [0-0.77]) 32 .
Recently, a 3D FT-CMR approach was introduced and permits a complete 3D interrogation of myocardial strain. Theoretically, a 3D-based strain approach is more suitable for the assessment of global myocardial deformation since 2D techniques are prone to through-plane loss of features in the third dimension. Due to LV contraction (with concomitant rotation) during systole that results in movement of some segments outside the pre-defined imaging plane, overestimation of the observed degree of muscle shortening may occur 33 . It is for this reason that 3D FT-CMR measurements result in lower absolute values. Hence, in the present study, we used a 3D FT-CMR approach to obtain global as well as regional longitudinal strain and to calculate the degree of relative "apical sparing".
Previous studies already showed that FT-CMR-based LV-GLS is significantly decreased in patients with CA compared to HCM 34 ; a finding that is confirmed by our present results. Prior to the introduction of strain analysis using FT-CMR, speckle tracking echocardiography (STE) was already widely used for years. Based on STE studies in patients with CA, the concept of relative "apical sparing" (with regard to longitudinal strain) was introduced 28 : As longitudinal strain impairment in CA predominantly affects the basal segments of the LV-whereas apical segments are affected to a much lower degree, the concept of relative "apical sparing" was established and initially shown to differentiate CA from other causes of LVH with a sensitivity of 93% and a specificity of 82% by Phelan et al. 28 . In addition to such diagnostic data, a prognostic value of STE-based assessment of "apical sparing" was described in case of CA 35 and this method was therefore even considered in current amyloidosis guidelines 36 . Since a good agreement was shown between STE-based and FT-CMR-based strain analysis, assessment of longitudinal strain was based on FT-CMR only in the present study 37 .
However, when Phelan et al. introduced the concept relative "apical sparing", they used a cut-off value for relative "apical sparing" of 1.0 in order to differentiate CA from HCM with the aforementioned good sensitivity and specificity 28 . In contrast, in a more recent study, Pagourelias et al. obtained a poorer sensitivity of 63% and similar specificity of 85% for the diagnosis of CA using an optimal cut-off value of 0.87 38 . In our present CMR study, an optimal cut-off of 0.89 for the parameter "apical-to-(basal + midventricular)" LV-RLS-ratio resulted in a 64% sensitivity and 80% specificity for the diagnosis of CA-confirming the data of Pagourelias et al. 38 .
In contrast to echocardiography, CMR allows not only to assess myocardial strain but also structural changes using conventional LGE techniques and novel parametric mapping approaches such as native T1-mapping and T1-mapping-derived ECV. Native T1-mapping has the advantage of detecting subtle changes in the intrinsic properties of the myocardium that may be indicative of an early stage of cardiac disease. Moreover, in pathologies with diffuse or uniform pattern, T1-mapping has a superior diagnostic accuracy than LGE-imaging that requires relative signal intensity changes and reference regions of normal myocardium in order to depict tissue abnormalities 39 40 . Moreover, the same group observed prognostic implications of T1-mappingbased analysis in CA patients-with ECV being independently predictive of death (hazard ratio [HR] 1.155 for each 3% increase; 95% CI 1.097 to 1.216; p < 0.001) in a multivariable Cox model that included (among others) N-terminal pro-B-type natriuretic peptide, LV-EF, LGE extent and LV mass 41 . Hence, both native T1 and ECV are highly sensitive non-invasive CMR parameters for the diagnosis of CA.
Surprisingly, an appropriate head-to-head comparison of the diagnostic performance of T1-mapping-based vs. FT-CMR-based parameters has not been performed in CA and HCM patients so far. To the best of our knowledge, this is the first study that provides a direct comparison of the diagnostic value between FT-CMR and T1-mapping in CA and HCM patients. In the present study, both native T1 and ECV substantially outperformed FT-CMR-based strain analysis with regard to the appropriate diagnosis of CA. Noteworthy, native T1 and ECV did not only demonstrate a substantially higher AUC based on ROC analyses compared to FT-CMR; in addition, ECV was an independent predictor of the presence of CA in multivariable regression analyses-in contrast to FT-CMR-based strain parameters. Hence, from a clinical point-of-view, strain-based assessment of relative "apical sparing" does not allow to safely rule out the presence of CA. Beyond strain measurement, novel CMR methods for myocardial tissue characterization such as T1-mapping and ECV measurement are required and should be routinely performed in patients with suspected CA and/or LVH of unknown origin. Moreover, a careful assessment of the diagnostic value of strain measurements is suggested in future recommendation and/ or guideline papers addressing the diagnosis of CA.
Limitations. Since CA is a rare disease, the size of our study group was limited 42 , however sufficient for the purposes of this study. In addition, CA consists of different subtypes dependent on the precursor protein.
In the present study, no distinction was made between the different types of CA (AL and ATTR) since the underlying mechanism of cardiac involvement is similar 43 . Furthermore, the limitations of T1-mapping and ECV measurement 44 as well as of FT-CMR are well documented 30  www.nature.com/scientificreports/ Gadovist (0.15 mmol/kg) for CMR imaging purposes and future studies need to show whether the licensed dose of 0.1 mmol/kg will be sufficient for tissue mapping purposes. However, based on previous data and according society recommendations 26 , our approach should be appropriate.

Conclusion
Native T1-mapping and ECV measurement are both superior to longitudinal strain measurement (with assessment of relative apical sparing) regarding the appropriate diagnosis of CA. Hence, cardiac work-up of patients with suspected amyloidosis should be based on multi-parametric CMR including T1-mapping and ECV measurement-and not limited to echocardiography-and/or CMR-based longitudinal strain analysis.

Data availability
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.