This study was conducted in accordance with the Declaration of Helsinki (2013 revision). Approval for this retrospective study was granted by the Ethics Committee on Scientific Research of our hospital; meanwhile, individual consent for a retrospective analysis was waived.

In this retrospective observational study, the magnetic resonance (MR) images of 89 consecutive patients who had undergone cardiac MR imaging for either clinically suspected or confirmed cardiac amyloidosis between January 2014 and November 2022 were acquired from the picture archiving and communication system (PACS). A diagnosis of AL-CA was established through a biopsy of subcutaneous fat or an involved organ demonstrating Congo red staining with apple green birefringence, detection of monoclonal protein in serum or urine, and/or identification of monoclonal plasma cells in the bone marrow. Furthermore, cardiac amyloidosis was defined as LV wall thickness exceeding 12 mm without another known cause, confirmed by echocardiography, diffuse late gadolinium enhancement on CMR imaging, or both [19]. According to the Boston University staging system, patients were classified as follows: (1) Stage I: Both BNP and cardiac troponin-I (cTnI) are below their thresholds (BNP 81 pg/mL and cTnI 0.1 ng/mL); (2) Stage II: One biomarker is above its threshold (BNP $\ge$81 pg/mL or cTnI $\ge$0.1 ng/mL), while the other is below; (3) Stage III: Both biomarkers are above their thresholds (BNP $\ge$81 pg/mL and cTnI $\ge$0.1 ng/mL) [20]. The exclusion criteria were a history of chemotherapy, poor image quality, and incomplete clinical data. The specific standards for the poor image quality were as follows: (1) Motion artifacts: Images with significant blurring or ghosting due to patient movement. (2) Low signal-to-noise ratio: Images where the myocardial borders were not distinguishable due to excessive noise. (3) Incomplete coverage: Images that did not fully recognize the entire left ventricular myocardium. The flow chart is shown in Fig. 1.

Fig. 1.

Flowchart of the patient selection process in the study according to inclusion and exclusion criteria. CMR, cardiac magnetic resonance; AL-CA, light-chain cardiac amyloidosis.

The primary endpoint of this study was defined as all-cause mortality, which was assessed through the patient’s hospital chart or telephone interviews conducted by an investigator who remained unaware of the patients’ clinical and cardiac MR imaging data. The follow-up period for the patients was extended until September 2023, and the duration between the date of cardiac MR imaging and the event of death was recorded as the time to event.

All patients underwent magnetic resonance imaging (MRI) scanning using a 3.0T scanner (Philips Ingenia (6.1.571.0, Philips Medical Systems, Best, The Netherlands), GE Discovery MR750 (DV24, GE Healthcare, Milwaukee, WI, USA), SIEMENS MAGNETOM Prisma (VE11E, Siemens Healthineers, Erlangen, Germany)) and a 1.5T scanner (SIEMENS MAGNETOM Aera (VE11C, Siemens Healthineers, Erlangen, Germany)) with a cardiac or abdominal coil. Images were acquired with retrospective electrocardiogram (ECG) gating during end-expiratory breath holding. SSFP cine images were obtained in consecutive short axes covering the LV and the long axis (two-, three-, and four-chamber views) according to the standardized protocol [21]. The detailed imaging parameters for each scanner are provided in Supplementary Table 1.

For analysis, CMR images were imported into CVI42 v5.14.2 software (Circle Cardiovascular Imaging Inc., Calgary, AB, Canada). Two experienced radiologists performed the measurements, each with 4 and 5 years of experience in CMR diagnosis. To analyze the short-axis images, the software automatically detected the endocardial and epicardial contours of the left ventricle during both the diastolic and systolic phases. Subsequently, manual adjustments were used to ensure accurate delineation layer by layer. General CMR parameters and CMR-FT analyses were conducted. The general CMR parameters were exported, including the left ventricular end-diastolic volume index (LVEDVi), left ventricular end-systolic volume index (LVESVi), and LVEF.

As for strain indicators, global strain parameters of the left ventricle, such as radial strain (RS), circumferential strain (CS), and longitudinal strain (LS), were obtained first. Secondly, the segmental strain parameters of the left ventricle were measured in different regions of the myocardium, including the basal, middle, and apical segments. The segmental strain parameters were further assessed in 16 segments of the myocardium using the American Heart Association’s 16-segment model [22].

Intra- and interobserver variabilities in strain values were evaluated in 20 randomly selected AL amyloidosis patients using the intraclass correlation coefficient. Interobserver variability was assessed by two independent investigators using the same image set. Intraobserver variability was evaluated on the identical image set by one investigator one month later.

Categorical variables are presented as the count and percentage; parametric continuous variables as the mean $\pm$ standard deviation (SD); non-parametric continuous variables as the median and interquartile range (IQR). Group differences in non-parametric continuous variables were assessed using the Mann–Whitney U test. Kaplan–Meier curves were utilized to estimate survival distribution over follow-up duration. The association of clinical, imaging, and serological parameters with outcomes was evaluated using both univariate and multivariate Cox proportional-hazards regression models. For Cox regression analysis, segmental strain values were transformed into categorical variables based on predefined cut-offs, where values above the cut-off were categorized as high and below as low. Receiver operating characteristic (ROC) curves were constructed to determine optimal cut-off values for endpoints, and the DeLong method was employed to compare the area under the curve (AUC). Statistical significance was defined as p 0.05. Statistical analyses were conducted using SPSS version 26 (IBM, Armonk, NY, USA) and GraphPad Prism 9 (GraphPad, San Diego, CA, USA).

Among the cohort of 89 patients, 80 received a diagnosis of AL-CA. One individual faced exclusion due to a history entailing chemotherapy, while four others were omitted owing to inadequate CMR image quality and incomplete clinical data. Consequently, the refined study population comprised 75 patients (50 men, with an average age of 56 $\pm$ 10 years). Over a median follow-up period of 34 months (ranging from 24 to 41 months), 16 patients (21%) experienced mortality events. The baseline clinical characteristics of the entire study population, categorized into survivors and non-survivors, are presented in Table 1. In the comparative analysis between these groups, survivors exhibited similar age, gender, body mass index (BMI), and blood pressure when compared to non-survivors (all p $>$ 0.05). Non-survivors demonstrated significantly elevated levels of Modification of Diet in Renal Disease (MDRD), BNP, cTnI, and Boston University (BU) stage when compared to survivors (p = 0.043, p 0.001, p 0.001, and p 0.001, respectively).

The comprehensive assessment of cardiac parameters, both general and strain-related, obtained from CMR cine is presented in Table 1. Survivors and non-survivors exhibited similarities in LVEDVi and LVESVi, with no statistically significant differences between the two groups (p = 0.154, p = 0.873, respectively). However, a notable distinction emerged in LVEF, which was significantly lower in the non-survivors than in the survivors group (p = 0.014).

Turning to the LV global strain parameters assessment, both global radial strain and circumferential strain exhibited a substantial change in the non-survivor group compared to the survivors (RS_global, p 0.001; CS_global, p 0.001). In terms of LV segmental strain parameters, significant differences were discerned between survivors and non-survivors in the basal segments of radial strain, circumferential strain, and longitudinal strain (RS_basal, p = 0.029; CS_basal, p 0.001; LS_basal, p = 0.023). Moreover, significant variations were noted in the middle segments of radial strain, circumferential strain, and apical segments of longitudinal strain between the two groups (RS_middle, p 0.001; CS_middle, p = 0.014; LS_apical, p = 0.002).

Fig. 2 illustrates the left ventricular strain parameters across the 16 myocardial segments defined by the American Heart Association (AHA). Noteworthy distinctions were identified in radial strains, specifically in the basal anteroseptal (p = 0.004), inferoseptal (p = 0.002), lateral (p 0.001), inferolateral (p 0.001), and anterolateral segments (p = 0.001). Furthermore, variations in radial strains were significant in the middle anterior (p = 0.011) and inferoseptal segments (p = 0.037). Examining circumferential strains revealed significant disparities in the basal inferoseptal (p = 0.020), lateral (p = 0.011), inferior (p = 0.003), anterolateral segments (p = 0.029), as well as in the middle anterior segments (p = 0.044). Additionally, longitudinal strains exhibited significant differences in the basal inferoseptal (p 0.001), middle anteroseptal (p = 0.034), apical inferior (p = 0.006), and apical lateral segments (p = 0.014). Representative CMR acquisitions of myocardial strain from survivors and non-survivors groups in our cohort are illustrated in Figs. 3,4.

Fig. 2.

The polar maps of myocardial strain values for the 16 segments were analyzed in subgroups of survivors (n = 59) and non-survivors (n = 16) with AL-CA. (A–C) illustrate the standard deviations of the mean 16 segmental strain values, including radial strain, circumferential strain, and longitudinal strain for survivors, while (D–F) illustrate the standard deviations for non-survivors. Statistical comparisons reveal significant differences († p 0.01 survivors versus non-survivors, ‡ p 0.05 survivors versus non-survivors). AL-CA, light-chain cardiac amyloidosis.

Fig. 3.

Quantitative measurements of myocardial strain analysis in a 63-year-old male patient with confirmed AL-CA who died after a survival time of 14 months. (A–D) Cardiac cine images illustrating the delineation of left ventricular endo- and epicardial contours (green represents epicardium, red represents endocardium) were obtained in short and long axes (two-, three-, and four-chamber views). (E–G) Myocardial strain curves of circumferential strain (CS), longitudinal strain (LS), and radial strain (RS). (H–J) The corresponding bull’s eye plots of the 16 myocardial segments. AL-CA, light-chain cardiac amyloidosis.

Fig. 4.

Quantitative measurements of myocardial strain analysis in a 56-year-old female patient with confirmed AL-CA who was alive after a survival time of 28 months. (A–D) Cardiac cine images illustrating the delineation of left ventricular endo- and epicardial contours (green represents epicardium, red represents endocardium) were obtained in short and long axes (two-, three-, and four-chamber views). (E–G) Myocardial strain curves of circumferential strain (CS), longitudinal strain (LS), and radial strain (RS). (H–J) The corresponding bull’s eye plots of the 16 myocardial segments. AL-CA, light-chain cardiac amyloidosis.

For survival analysis, two distinct approaches were employed (Table 2). Initially, clinical information and global strain parameters were considered, followed by a specific examination of the 16-segment strains.

The univariate Cox regression analysis identified several predictors of all-cause mortality, including BNP levels, BU stage, LVEF, and various global strain indices from the CMR assessments. In the subsequent multivariate Cox analysis, which included BNP, BU stage, LVEF, RS_global, CS_global, and LS_global (all p 0.05), both BNP (hazard ratio: 1.001; 95% CI: 1.001, 1.002; p 0.001) and RS_global (hazard ratio: 0.225; 95% CI: 0.057, 0.887; p = 0.033) remained significant independent predictors of all-cause mortality.

Expanding the analysis to include the 16-segment strain measurements of RS, such as RS_{bas-inferoseptal}, RS_bas-inferor, RS_{bas-inferolateral}, RS_{bas-anterolateral}, RS_mid-anterior, and RS_{mid-inferoseptal}, further underscored their individual importance as predictors of all-cause mortality in AL amyloidosis patients. In the multivariate Cox analysis incorporating predictors with p 0.05 from the univariate Cox regression, both BNP (hazard ratio: 1.001; 95% CI: 1.001–1.002; p 0.001) and RS_bas-inferior (hazard ratio: 0.168; 95% CI: 0.035–0.803; p = 0.025) emerged as significant independent predictors of all-cause mortality.

In the ROC curve analysis (Fig. 5), BNP demonstrated the highest discriminatory ability (AUC = 0.901, 95% confidence interval (CI): 0.811, 0.958) when comparing survivors and non-survivors. RS_global exhibited an AUC of 0.681 (95% CI: 0.563, 0.784), while RS_bas-inferior had an AUC of 0.786 (95% CI: 0.676, 0.872). Significant differences were found between the fields under the ROC curves for BNP and RS_global and BNP and RS_bas-inferior (p = 0.002, p = 0.029, respectively). Stratification was applied to RS_global and RS_bas-inferior, with cut-off values of 24.97% and 20.74%, respectively. Kaplan–Meier survival curves were constructed for RS_basal and RS_bas-inferior subgroups, revealing notable differences in survival probability. A low RS_global level (below the cut-off value of 24.97%) and a low RS_bas-inferior level (below the cut-off value of 20.74%) were associated with an increased risk of death (Fig. 6, log-rank p = 0.013, log-rank p 0.001, respectively).

Fig. 5.

Receiver operating characteristic (ROC) curves for B-type natriuretic peptides (BNP), global radial strain (RS_{𝐠𝐥𝐨𝐛𝐚𝐥}), and radial strain in the basal inferior segment (RS_bas-inferior) for the endpoint.

Fig. 6.

Kaplan-Meier curves for event occurrence stratified by radial strain parameters. Kaplan–Meier estimations of event occurrence over time, analyzed in relation to two specific parameters: global radial strain (RS_global) with an optimized cut-off of 24.97% (A) and radial strain in the basal inferior segment (RS_bas-inferior) with an optimized cut-off of 20.74% (B) for the endpoint.

Inter- and intraobserver variabilities of strain values were analyzed (Table 3), and reproducibility of the LV strain was relatively desirable. The LV strain was highly reproducible, exhibiting a high intraclass correlation (inter- and intraobserver intraclass correlation coefficient $>$0.80).