Introduction
Amyloidosis is a multi-system disease that is caused by the synthesis and accumulation of unstable and misfolded proteins leading to a loss of normal tissue architecture and function [
1]. While almost every organ can be affected by amyloid deposition, the prognosis of patients is mainly determined by the occurrence and extent of myocardial involvement [
2‐
4]. There are various types of amyloid that differ in the source and nature of the misfolded precursor protein. Of those, light chain associated amyloidosis (AL) and transthyretin amyloid amyloidosis (ATTR) are the two main types of amyloidosis that affect the heart. The infiltration of amyloid to the heart leads to myocardial thickening and diastolic dysfunction resulting in heart failure.
For the assessment of cardiac amyloidosis (CA), cardiac imaging plays a fundamental role [
5]. Besides myocardial thickening, echocardiography often shows a reduction of global longitudinal strain, which typically spares the apical segments [
6]. Non-invasive characterization of the myocardial tissue of amyloidosis patients is the strength of cardiovascular magnetic resonance (CMR). In particular, the extent and distribution of late gadolinium enhancement (LGE) has been shown to provide independent information on outcome even after adjustment for known prognostic factors [
7]. CA is also an ideal application for newer mapping techniques as it leads to excessively high T1- and extracellular volume (ECV) [
8,
9]. Notably, in AL CA patients, ECV is a proven predictor of mortality with an up to four-fold increased likelihood of death for ECV > 0.45 over a median follow-up period of 23 months [
10]. Both LGE and ECV, however, rely on the application of gadolinium-based contrast agents (GBCA), which is often limited in amyloidosis patients due to severely reduced renal function.
Recently, CMR diffusion tensor imaging (DTI) emerged as a promising method for the determination of myocardial fiber orientation [
11‐
13] and tissue characterization without the need for GBCA. It allows to investigate water diffusion within the tissue and to derive additional scalar metrics, such as mean diffusivity (MD) and fractional anisotropy (FA) for the quantification of structural integrity [
14,
15]. Alterations of cardiac microstructure have been described in patients with dilated cardiomyopathy (DCM), myocardial infarction and hypertrophic cardiomyopathy (HCM) [
16‐
18]. In CA, however, the effect of the double harm resulting from the combination of amyloid infiltration with direct cardio-toxicity of abnormal proteins [
19] on the myocardial structure and its potential impact on diffusion properties is yet unknown.
The objective of this study was to determine the consequences of amyloid infiltration on myocardial microstructure in-vivo using CMR DTI and to investigate their effect on myocardial function. Scalar diffusion and relaxation metrics (MD/FA and native T1/ECV, respectively) were correlated in CA patients and controls. In addition, directional diffusion information was assessed to reveal changes in myofiber orientations and compared to parameters of left ventricular (LV) function.
Methods
Study design
Between March 2017 and June 2018, 12 patients with AL or ATTR CA (63 ± 11 years, 3 female), referred to the local outpatient clinic, and 10 healthy, age-matched controls (62 ± 11 years, 4 female) were prospectively enrolled. Exclusion criteria were cardiovascular disease other than CA, kidney failure with estimated glomerular filtration rate (eGFR) < 30 ml/min/1.76 m
2 and the standard exclusion criteria for CMR [
20]. Height and weight of all study subjects were recorded to calculate the body surface area (BSA). All patients had histologically proven (positive Congo red staining of endomyocardium, abdominal fat, kidney, rectum or bone marrow biopsies) amyloidosis and echocardiographic or CMR findings typical for cardiac involvement. Imaging was performed on a clinical 1.5 T system (Achieva, Philips Healthcare, Best, The Netherlands) equipped with a 5-channel cardiac receiver array. Prior to imaging, written informed consent was obtained from all subjects. The study protocol was approved by the ethics committee of the canton of Zurich; it allowed the administration of GBCA only in the CA patient group.
CMR data acquisition
Cardiac function was assessed by a contiguous stack of balanced steady-state free precession (bSSFP) short-axis cine images covering the entire LV and bSSFP long-axis cine images in 2-chamber, 3-chamber and 4-chamber orientations. According to the cine images, systolic quiescent time points were determined on a per subject basis.
Diffusion-weighted imaging was performed with a single-shot spin-echo sequence utilizing second-order motion compensated diffusion sensitizing gradients in combination with an echo planar imaging (EPI) readout [
21,
22]. The sequence was electrocardiogram (ECG)-triggered to mid-systole systole (65% of peak systole) and three imaging planes were placed in short-axis view orientation at apical, mid-ventricular and basal level LV using a reduced field-of-view (FOV) technique [
23]. Diffusion weighting was encoded along nine and three directions with a b-value of 450 and 100 s/mm
2, respectively [
24]. Data acquisition was performed during free breathing with respiratory navigator-based slice tracking [
25]. The imaging parameters are listed in Table
1.
Table 1
CMR Scan Parameters
Spatial Resolution [mm3] | 2.5 × 2.5 × 8 | 1.2 × 1.2 × 8 | 1.6 × 1.6 × 10 |
FOV [mm2] | 230 × 105 | 300 × 300 | 360 × 460 |
TR | 3 beats | 2.5 ms | 3.5 ms |
TE [ms] | 76 | 1 | 1.7 |
Flip Angle [°] | 90 | 35 | 15 |
Number of Averages | 12 | 1 | 1 |
Respiratory Mode | Free Breathing | Breath Hold | Breath Hold |
Approx. Scan Time [min.] | 8–10 min | 1–2 min | 1 min |
T1 mapping was performed using a modified Look-Locker inversion (MOLLI) sequence with a total of eight T1 weighted images split into a set of five and three images acquired after inversion, separated by three heartbeats of recovery (5-(3)-3 scheme) [
26]. The imaging planes were at the same slice location as the DTI scan. In all subjects, native T1 maps were acquired. For ECV calculation, T1 mapping was repeated in the patient group 20 min after a bolus injection of 0.2 mmol/kg gadolinium-based contrast agent (Gadovist, Bayer Schering, Berlin, Germany). The hematocrit was measured in all patients within 2 h of CMR. LGE images were acquired with an inversion-recovery sequence approximately 10 min after the bolus injection.
CMR data analysis
Standard measurements of LV size and function including LV ejection fraction (LVEF), left ventricular mass index (LVMi) and intraventricular septal wall thickness (IVS) were obtained using the GTVolume software package (GyroTools LLC, Zurich, Switzerland). The LV end-systolic global longitudinal strain (GLS) and global circumferential strain (GCS) were calculated using a certified CMR feature tracking evaluation software (2 D CPA MR, Cardiac Performance Analysis MR Version 4, TomTec Imaging Systems, Unterschleissheim, Germany) on the long- and short-axis bSSFP images.
Prior to diffusion tensor calculation and T1 fitting, the individual MOLLI and DTI images were registered using a non-rigid groupwise image registration method [
27] for compensation of residual motion. DTI analysis was performed on MD, FA, helix angle (HA), transverse angle (TA) and absolute E2A sheet angle [
16,
28]. A detailed description of the DTI parameters is provided as Additional file
1. For calculating the HA slope, a linear regression of the transmural course of the helix angles between the endocardium and epicardium was performed, excluding boundary pixels at the endo- and epicardium. Angular diffusion metrics were evaluated in four radial and circumferential sectors of the left ventricle and averaged across all three slices. Mean LV signal-to-noise ratio (SNR) of the b = 100 s/mm
2 image was determined in both patient groups for quality control. The scalar diffusion metrics MD and FA as well as native T1 and ECV, were evaluated in a single septal region-of-interest (ROI) on the midventricular short-axis slices in accordance with current clinical recommendations [
29]. ECV was computed based on the collected haematocrit, native and enhanced T1 values in the myocardium and the blood pool.
Statistical analysis
All data are expressed as mean ± SD. Statistical analysis was conducted using MedCalc software (MedCalc 17.9.7, MedCalc Software bvba, Ostend, Belgium). Differences in diffusion and relaxation metrics between patients and controls were assessed by unpaired two-tailed t-tests. To determine statistical associations between parameters, Pearson correlation analyses were performed. In addition, the diagnostic accuracy of native T1, MD and FA for detecting cardiac amyloidosis compared to healthy controls was examined by ROC analysis. A p-value less than 0.05 was considered statistically significant.
Discussion
This study assessed the value of cardiac DTI for the evaluation of patients with CA and investigated the effects of amyloid deposition on myocardial microstructure.
The scalar DTI metrics of MD and FA were found to be significantly altered by CA. In comparison with native T1, which was also elevated in CA patients in agreement with previous studies [
10,
30], MD and FA exhibited a comparable AUC to identify CA patients indicating potential for similar diagnostic accuracy. The correlation between MD and native T1 suggests that the pathomechanism of amyloid infiltration affects both parameters to a similar extent. While native T1 is a useful diagnostic marker for CA, recent studies have shown that it is inferior to ECV as an independent predictor of prognosis in AL-amyloidosis as well as ATTR-amyloidosis patients [
30,
31]. However, the significant number of amyloidosis patients suffering from renal failure stresses the need for a gadolinium-free substitute for ECV with potential prognostic value. The robust correlation between FA and ECV found in the CA patients justifies expectations, that FA may have potential to serve as such a substitute for ECV. This warrants future trials to investigate the diagnostic accuracy of FA and MD for amyloidosis in particular against other cardiomyopathies with a HCM phenotype. The existence of a tight functional relation between FA and ECV is also suggestive from a microstructural perspective, since the scalar CMR DTI parameters describe the range and anisotropy of myocardial water movement, which is mainly determined by the tissue-specific presence of spatially ordered cellular compartments within the myocardium [
32]. Our findings are supported by previous studies, which have demonstrated that quantitative measures of diffusivity derived from CMR DTI capture similar properties as LGE and ECV in myocardial infarction and HCM [
33,
34].
Time efficiency is an essential criterion for clinical practice under the economic constraints of healthcare systems. From this point of view, measurement durations of approximately ten minutes seem unreasonably long and insufficient for daily routine in CMR imaging. However, the presented sequence acquired three slices, two of which, namely the basal and apical slice were only used for the determination of myocardial microarchitecture. The MD and FA values were only assessed in the mid-ventricular slice as recommended in the current consensus document [
29]. Limiting the acquisition to a single mid-ventricular slice would reduce the measurement time by approximately 25% assuming a 1.5 T CMR system and a heart rate of 60 beats per minute. In addition, replacing LGE and ECV with DTI would also eliminate the 10–15 min waiting period before LGE and ECV [
35,
36] imaging and could ultimately lead to a saving of time.
Our investigation of myocardial architecture provided new insights into the microstructural alterations due to interstitial amyloid deposition. The most prominent observation of significantly increased absolute E2A sheet angle is the structural equivalent of a more systolic conformation of the myocardial sheetlets. The correlation between E2A sheet angle and GLS may indicate that, at a microstructural level, a pronounced systolic sheetlet configuration might be associated with a limited ability to rotate to a relaxed conformation in diastole. Similar observations have been made in HCM patients [
17], indicating that despite fundamentally different pathogenesis, comparable structural alterations may cause the primary diastolic heart failure observed in both diseases. The difference in HA between health and disease, however, has not yet been described. The significantly lower HA slope of CA patients results in a reduced transmural HA gradient with a less longitudinal orientation of the myocardial fibers. The strong correlation which we observed between the HA slope and the GLS indicates, that the more circumferential myofiber orientation may be the microstructural determinant for the loss of longitudinal function in amyloidosis patients. This finding seems not to be a common feature of all hypertrophic phenotype cardiomyopathies as Nielles-Vallespin et al. described similar HA configuration in HCM patients and healthy controls [
17]. In agreement with the hypothesis of the HA slope being a determinant for longitudinal function, a recent study by Pagourelias et al. demonstrated that the GLS of CA patients is significantly lower compared to HCM patients with identical myocardial wall thickness [
37]. In addition to the microstructural alterations, the direct cardio-toxicity of amyloid deposition may also contribute to a loss of myocardial function in CA patients [
19].
It remains unclear whether the relation between HA slope and GLS is linear over a wider range. A previous study by von Deuster et al. [
16] in DCM patients reported similar HA slope in controls and patients with mildly reduced LVEF, but two patients with advanced heart failure showed reduced GLS and increased HA slope. In synopsis, these findings support the hypothesis of a U-shaped relation between HA slope and GLS with an ideal HA slope around 1°/%
tmd and impaired longitudinal function caused by deviations to significantly higher and lower HA slope values. The different constellations of DTI parameters observed in HCM, DCM and amyloidosis raise the hope that future research may identify disease specific patterns of DTI parameters, which would be an advantage over rather unspecific parameters such as native T1 that is elevated in a multitude of cardiac diseases. The presence of subtype specific patterns of the DTI parameters for AL- and ATTR-amyloidosis should be investigated, in particular in comparison to nuclear imaging techniques where radiotracers used for technetium-labeled bone scintigraphy are found to have high sensitivity for ATTR [
38], while
18F-florbetaben-PET has recently been shown to have a high sensitivity for AL-amyloidosis [
39].
In contrast to HA and E2A, the transverse angle does not seem to provide relevant insights into disease specific pathologies as it is distributed around 0° in health and disease. However, the elevated standard deviation of TA that we observed in amyloidosis patients has previously already been observed in DCM patients [
16] and may therefore be a surrogate for myocyte disarray of any cause. A bias of the TA distribution due to low SNR, which has been observed in previous studies [
24], could be excluded in our work since SNR values were not different between the amyloidosis and control cohorts.
Limitations
A limitation of the present work is the small sample size of CA patients with different types of the disease. In particular the patient number is insufficient to investigate potential effects of the different pathophysiologies of AL- and ATTR-amyloidosis on the myocardial microstructure. In addition, the lack of GBCA application in healthy controls, due to constraints from ethical approval constitutes a limitation. However, despite those limitations, our patient population already showed promising diagnostic properties of CMR DTI based parameters warranting future research in a larger population of CA patients. In contrast to previous studies using STEAM based DTI methods [
40], this study was limited to investigate the mid-systolic configuration of the myocardium and did not assess a diastolic state.
Conclusion
The scalar CMR DTI parameters MD and FA in CA patients are significantly different from healthy controls, allowing to assess the amyloid induced microstructural disarray without the use of GBCA. The major differences in myofiber orientation are a significantly increased E2A sheet angle and a reduced HA slope in CA patients. Finally, the strong correlation between HA slope and GLS indicates, that the more circumferential orientation of the myofibers may provide the rationale for the reduction of CLS that occurs in CA.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.