4D Flow Component Reveals Left Ventricular Function In Atrial Fibrillation


 Purpose: Applicating cardiovascular magnetic resonance (CMR) 4D flow to evaluate left ventricular systolic and diastolic function in atrial fibrillation (AF).Methods: In this study, from May 2021 to October 2021, 26 AF patients and 15 healthy participants were recruited and underwent multiparametric CMR and echocardiogram scans before discharge. The CMR protocol incorporated an assessment of 4D flow, cardiac function. Notably, the AF patients maintained irregular heart rhythm during the CMR scan. The 4D flow components were compared with echocardiogram results.Results: AF patients had a lower proportion of direct flow (33.5% (8.43; 47.5) in AF vs. 69.1% (63.4; 74.4) in healthy), higher delayed ejection (24.1±12% in AF vs. 14.7±6.8% in healthy), retained inflow (32.5% (24.6; 36.9) in AF vs. 14.5% (12.5; 18.2) in healthy) and residual volume (4.73% (1.47; 15.0) in AF vs. 0.62% (0.46; 1.62) in healthy). A high correlation was observed between flow EF and CMR EF (R=0.78, P<0.001) in AF patients. The mean bias among the methods was higher for flow EF than CMR EF (24.86%). Multivariable linear regression showed that the correlation between retained inflow and E/e’ (β=0.152, P=0.048) remained significant when adjusted for confounders. Direct flow (β= -0.170, P=0.038) and retained inflow (β=0.350, P=0.024) significantly correlated with the Minnesota Living with Heart Failure Questionnaire score. Conclusion: CMR 4D flow component revealed a significantly different flow pattern in AF patients compared with healthy participants and provided de novo flow biomarker indicating LV systolic function (direct flow and flow EF) and diastolic function (retained inflow).


Introduction
Worldwide, atrial brillation (AF) is the most common arrhythmia in adults, associated with high morbidity and mortality [1]. A critical biomarker-left ventricular ejection fraction (LVEF), predicts the survival of patients with AF and the effect of catheter ablation [2,3]. Conventionally, LVEF is derived from echocardiography in clinical practice [4]. Although echocardiographic grading scales are well established, allowing the evaluation of left ventricular diastolic function in sinus rhythm [5,6], assessing diastolic abnormalities in AF remains clinically challenging and often disregarded due to its irregular beating pattern.
Cardiac magnetic resonance (CMR) is the gold standard for the evaluation of cardiac structural and functional abnormality [7]. An emerging strategy called 4D ow CMR enables the acquisition of complexing blood ow in three directions simultaneously within a period of time [8]. It obviates the need in conventional 2D phase-contrast CMR ow, which manually aligns single velocity-encoding direction with target ow signals. The previous study had proved that 4D ow evaluated left atrium (LA) and left atrium appendage blood ow dynamics; results correlated strongly with transesophageal echocardiography velocities (r= 0.41, P <0.05) and stasis (r= -0.39, P <0.05) in the context of AF rhythm [9]. However, few studies reported ventricular function evaluation in AF rhythm through 4D ow. A technique named ow component showed much advanced practical meaning in normal and disease states [10,11]. It enables visualization of three multidimensional images of ow that allow quanti cation of intraventricular blood ow; it may re ect changes in left ventricle (LV) con guration, myocardial function, and pressure distribution within the disease, eventually associated with LV function [12].
This study aims to evaluate the e cacy of 4D ow MRI and ow component in the context of atrial brillation and explore its association with left ventricular systolic and diastolic function.

Participant
Patients admitted to the hospital from May 1 st, 2021, to October 20 th, 2021, were enrolled in this research. Patients were eligible if they were diagnosed with atrial brillation. CMR scan was performed when patients maintained AF rhythm. The exclusion criteria included younger than 18 years old, inability to be placed in a magnetic resonance image (MRI) scanner due to body mass, pacemaker, and severe renal abnormality (glomerular ltration rate < 30 mL/min/1.73 m 2 ), hemodynamically unstable. Persistent AF was de ned as AF heart rhythm lasting longer than seven days; Paroxysmal AF was de ned as AF rhythm lasting shorter than seven days; Additionally, healthy participants were recruited as reference. All subjects provided written informed consent. The institutional review board approved the study.

Echocardiogram And Clinical Data Collection
The echocardiogram was performed on a EPIQ7 system with an X5-1 matrix transducer (Philips Medical Systems, Andover, MA). Echocardiogram parameters, including LA dimension (de ned as the largest diameter of LA in parasternal long-axis view), echo EF (Simpson bi-plane mode using apical fourchamber and two-chamber view) was measured. E/e', E, Septal e, tricuspid regurgitation velocity, LA volume index was measured according to recommendations of the American Society of Echocardiography and the European Association of Echocardiography. CHA2DS2-VASc score was calculated to assess the stroke risk level. HAS-BLED score was calculated to assess the bleeding risk level. Laboratory results in AF patients were acquired before discharge, including serum creatine, alanine aminotransferase (ALT), hemoglobin (Hb), brain natriuretic peptide (BNP), N-terminal pro-B type natriuretic peptide (NT-proBNP). The Minnesota Living with Heart Failure Questionnaire (MLHFQ) is used to evaluate the quality of life and symptom burden of patients [13,14].

Image Analysis
Two experienced clinicians analyzed the CMR data using Cvi42 version 5.13.2 (Circle Cardiovascular Imaging Inc., Calgary, Canada). LA and LV functions were analyzed o ine. Left ventricular (LV) volumes (LV end-diastolic volume (EDV), LV end-systolic volume (ESV), stroke volume (SV), and cardiac output (CO)), LV mass index, and CMR EF (derived from LV) were measured and analyzed by standard volumetric techniques through short-axis cine images. Cardiac strain analysis included global longitudinal strain (GLS, %), global radial strain (GRS, %), and global circumstance strain (GCS, %). GLS was derived from two-, three-, and four-chamber views, whereas GRS and GCS parameters, were derived from the short-axis stack.
The 4D ow datasets were imported into the Cvi42 version 5.13.2 (Circle Cardiovascular Imaging Inc., Calgary, Canada) for further analysis. Images were corrected for background offset errors and velocity aliasing artifacts. For semiautomatic (valve tracking) analysis, the three-chamber view of cine images from the study was used as the reference for aortic valve and mitral valve tracking. The valve contours on each phase of the 4D Flow series were corrected manually. Isovolumetric relaxation phase was set as the interval between closing of aortic valve and opening of mitral valve. The position of pathlines at endsystole divides them into four functional ow components as described previously [15,16]: '(1) direct ow: blood ow that enters and exits the LV in the analyzed cardiac cycle; (2) retained in ow: blood ow that enters the LV but does not exit during the analyzed cycle; (3) delayed ejection ow: blood ow that starts within the LV and exits during the analyzed cycle; and (4) residual volume: blood ow that remains in the LV for at least two cardiac cycles' (Figure 1). Each component volume was calculated as a proportion of the total end-diastolic volume. Ejection fraction derived from 4D ow, namely ow EF, was calculated as the equation displayed: ow EF = direct ow + delayed ejection.

Statistical analysis
Categorical and consecutive data were presented as number (%), mean ± standard deviation (normal distribution), or median ± quartile (non-normal distribution). Differences between means were tested by the unpaired t-test or Kruskal-Wallis test as appropriate. Pearson correlation was used to assess the correlation between variables. Statistical signi cance was de ned as P < 0.05. Univariate and Multivariable linear regression was carried out to investigate the association of ow component with diastolic function and MLHFQ. Statistical analysis was performed using the R package.

Result
Baseline characteristics A total of 41 subjects were recruited for our study, including 26 AF patients and 15 healthy volunteers. One patient and one healthy participant was excluded due to poor 4D ow data quality. Baseline characteristics including demographics, clinical characteristics, echocardiogram data, and CMR data were listed in Table 1 and

Changes in ow component
Flow visualizations in the systolic phase, the diastolic phase, and the interval between the diastolic and the systolic phase were shown in Figure 2  Subgroup analysis was conducted in AF and healthy participants seperately. Pearson analysis was performed to assess the correlation between direct ow, ow EF and CMR EF. For direct ow and CMR EF, a borderline (R=0.52, P=0.058) and signi cant positively correlation (R=0.72, P<0.001) were observed in healthy volunteers and AF patients respectively. For ow EF and CMR EF, high correlation was both observed in healthy volunteers (R=0.59, P=0.028) and AF patients (0.79, P<0.001).  (Table 3).

Flow component and symptom burden
The symptom burden and quality of life was evaluated with MLHFQ. The mean (SD) MLHFQ score of AF patients was 12.6±8.9 (Table 1)

Discussion
The results of our study demonstrated several important ndings regarding the diagnostic value of 4D ow component measurements in patients with AF. First, we found signi cantly reduced direct ow and added delayed ejection, retained in ow, and residual volume in AF patients compared with healthy participants. These results suggested that patients with AF had speci c ow patterns, which were important indicators of ventricle function, both diastolic and systolic. Second, two new biomarkers, namely ow EF and direct ow derived from 4D ow correlated with CMR EF and biomarkers derived from CMR, echocardiogram, and serum sample. Third, we found that retained in ow related to diastolic function (E/e') and remained signi cant after adjusting for confounders. Finally, direct ow and retained in ow associated with symptom burden and quality of life in AF patients.

Flow Component In Systolic Function
Cardiac magnetic resonance is generally considered as the gold standard to assess the systolic function by traditionally measuring the parameters such as LVEF, myocardial strain, and stroke volume [17].
However, accurate assessment of systolic function in patients with atrial brillation is still challenging due to the beat-to-beat irregularity and elevated ventricular rate. Unlike traditional evaluation technologies, 4D ow CMR enables to comprehensively evaluate the intracardiac ow in three directions throughout the cardiac cycle [18].
It is also reported that 4D ow technology may potentially measure cardiac function among atrial brillation patients. In the previous study, Kim et al.
[19] used 4D ow CMR to compare hemodynamics in 30 healthy controls and 50 paroxysmal atrial brillation patients. In Kim's study, compared with the control group, the ratio of direct ow was lower in the paroxysmal atrial brillation group (44.5 ± 11.2% vs. 50.0 ± 12.2%), while the delayed ejection was higher (21.6 ± 5.6% vs. 18.6 ± 5.7%). In our study, ow components signi cantly differed between the atrial brillation group and the healthy controls, with the lower direct ow and a higher ratio of the other three. However, it is worth noting that Kim et al. performed the analysis in the paroxysmal atrial brillation patients with sinus rhythm. In our study, we carried out the 4D ow CMR among the patients maintaining AF rhythm during the CMR scan, which may be helpful to gain insight into the hemodynamics change during AF. Moreover, we also found that the ow EF strongly correlates with the other con rmed systolic-related parameters, CMR EF and SV. Our analysis indicated that 4D ow component analysis, especially this new parameter ow EF, may serve as a biomarker of LV systolic function in AF patients.
Flow EF seemed much higher than clinical de ned EF, which might originated from ow separation and visualization method. Our measurement in healthy participants (Direct ow 69.  [25] reported that peak spatially integrated vorticity derived from 4D ow might re ect right ventricular diastolic function. In the present study, we found a signi cant correlation between retained in ow and diastolic function parameters (E/e', Septale e', LA volume index) in all subjects; in the AF group, this correlation remained signi cant after adjusted for confounders including demographics and cardiac function. It is worth mentioning that retained in ow describes the blood ow dynamics in the diastolic phase [12]. Hence the association between retained in ow and diastolic function was expected. Moreover, advanced methodology, including kinetic energy, has the potential to analyze cardiac function [20,26].

Age And Sex Difference In Flow Component
Age and sex may introduce a difference in ow component analysis. We had relatively balanced sex and body mass index; however, age was not perfectly matched. We introduced multivariable linear regression to balance these demographic characteristics, including age and sex, and we witnessed similar results. A multicenter study of CMR 4D ow demonstrated that sex differences in retained in ow, residual volume was evident; however, age difference regarding ow components was not evident[20].

Limitation
This study had limitations. First, it was a single-center, small sample study, with an age difference between the AF and healthy groups. Hence, we adopted multivariable linear regression to adjusted age confounders and compared with previous study [11] to explore the age in uence on ow component. Small sample size limited linear regression model effect. Second, 4D ow was sensitive to noise, arrhythmia, and motions; besides, a respiratory navigator was not applied in acquiring these data, which may cause artifacts. We excluded participants with low image quality. Our CMR protocol followed the recommendations, of which the stability of 4D ow was con rmed [8]. Third, cine images derived from CMR were accompanied by artifacts due to arrhythmia, in uencing the accuracy of LV function evaluation. Therefore, we adopted echocardiogram results parameters as a supplement.

Conclusion
In conclusion, the CMR 4D ow component revealed a signi cantly different ow pattern in AF patients than healthy participants and provided a de novo ow biomarker indicating LV function. 4D ow-derived EF may re ect LV systolic function, and retained in ow may be associated with diastolic function in AF patients. These biomarkers may serve as a meaningful tool, enabling observation of dynamic changes in AF patients and symptom burden. At the same time, CMR 4D Flow visualizes more versatile, comprehensive, and minor modi cations in intra-cardiac ow than echocardiogram (regardless of 2D or    Figure 1 Illustration of ow component LV blood volume is separated into four components. Direct ow is a part of in ow that enters ventricle during diastole, exits during systole (green line). Retained in ow is another part of in ow that remains in the ventricle during the systole (yellow line). Delayed ejection is a part of LV volume during diastole that exits during systole (blue line). Residual volume stays in the ventricle for more than one cycle (red line).
LV left ventricle.

Figure 2
Flow component visualization of a AF patient and a healthy participant Images of ow components in peak systole, peak diastolic and interval between them, captured from a healthy participant (top row) and an atrial brillation patient (bottom row). Direct ow was marked with green lines. Retained in ow was marked with yellow lines. Delayed ejection was marked with blue lines.
Residual volume was marked with red lines.  Direct ow and owEF compare with LVEF in healthy volunteers and AF patients A&B, correlation of direct ow and LVEF. C&D, correlation of ow EF and LVEF. Direct ow and ow EF was derived from 4D ow. LVEF was derived from cine images EF ejection fraction. LV, left ventricle; EF, ejection fraction.