A total of 49 patients with BAV-RL and 25 BAV-RN with AAo diameters ≤ 50 mm and no severe valvular disease (aortic regurgitation ≤ III, maximum aortic valve velocity <3 m/s by echocardiography) were consecutively and prospectively recruited between June 2014 and December 2015 at the Hospital Universitari Vall d'Hebron (Barcelona, Spain). Inclusion criteria were as follows: age >18 years, no connective tissue disorders, no aortic coarctation or other congenital heart diseases, no previous aortic surgery or aortic valve replacement, and no contraindication for CMR. A total of 48 HVs matched for sex, weight, height, body surface area (BSA), and stroke volume were also included. The local ethics committee approved the study, and informed consent was obtained from all participants.

4D-flow CMR data were obtained in a clinical GE 1.5T Signa scanner (GE Healthcare, Waukesha, WI, USA) using the Vastly undersampled Isotropic Projection Reconstruction (VIPR) technique (28, 29). The volumetric acquisition included the entire thoracic aorta and was performed with retrospective ECG-gating during free-breathing and without administration of an endovenous contrast agent. Acquisitions were made using the following parameters: field of view (FOV) 400 mm × 400 mm × 400 mm, voxel size of 2.5 mm × 2.5 mm × 2.5 mm, flip angle 8°, repetition time 4.2–6.4 ms, echo time 1.9–3.7 ms, and velocity encoding 200 cm/s. This dataset was reconstructed with a temporal resolution that ranged between 21–32 ms.

The two-dimensional balanced steady-state free precession (bSSFP) cine CMR images were used to assess BAV morphotype and aortic diameters as described (16). Briefly, the three aortic root cusp-to-commissure diameters were measured using double-oblique cine images at the level of the aortic root at end-diastole, and the maximum value was retained for the analysis. Similarly, AAo diameter was measured at end-diastole by double-oblique cine CMR at the level of the pulmonary artery bifurcation. To determine the presence of aortic root or ascending dilation, aortic diameters were adjusted with a logarithm transformation to set the z-score for both sinuses (zsinus) and AAo (zAAo) accounting for sex, age, and BSA as described by Campens et al. (30). Using a z-score cutoff value for the definition of aortic dilation of two standard errors of the estimate, patients were categorized according to the aorta segment predominantly or exclusively involved in dilation according to Della Corte's classification (31). Thus, patients were classified as non-dilated (zsinus ≤ 2 and zAAo ≤ 2), root-dilated morphotype (zsinus>2 and zsinus>zAAo), or AAo-dilated morphotype (zAAo>2 and zAAo>zsinus). Given the small number of patients (seven patients) with rood-dilated morphotype that we have, the groups of patients with BAV were divided only into two morphotypes categories, BAV-Non-AAoD (where we include 18 non-dilated patients and seven root-dilated patients) and BAV-AAoD (where we include the 49 patients with dilatation of the ascending aorta).

The hemodynamic parameters and the diameters in the entire thoracic aorta were obtained based on a finite-element method described previously (32–36) (Figure 1A). To provide a ground for location-specific comparisons, 16 regions (Figure 1D) were semi-automatically selected. Those regions were delimited by anatomical landmarks, regions 1 to 4 = AAo (between the Valsalva level and the brachiocephalic trunk), regions 5 to 8 = aortic arch (AArch, between the brachiocephalic trunk and the isthmus level), regions 9 to 12 = proximal descending aorta (pDAo, between the isthmus level and the level of Valsalva), and regions 13 to 16 = distal descending aorta (dDAo, between the Valsalva level and the diaphragmatic level).

In each region, the mean value of 17 hemodynamic parameters and the aortic diameter were calculated (Figures 1B,C) for each cardiac phase. Peak systolic values were retained for comparison [averaged at peak systole using 1 time-frame before and two time-frames after to reduce noise in the data (16)], except for regurgitation fraction (35) and oscillatory shear index (OSI) calculated along the entire cardiac cycle. The detailed information about the quantification of each parameter is provided in Supplementary Figures S1–S8.

The global statistical differences between healthy volunteers and patients with BAV were evaluated using a multivariate analysis of variance (MANOVA), with Pillai's trace statistic, whereas the local statistical differences were evaluated using the Mann–Whitney U test. To determine which parameters were more relevant to classify BAV morphotypes and phenotypes with respect to healthy volunteers, the receiver operating characteristic (ROC) curves and the minimum redundancy maximum relevance (MRMR) algorithm were used. Moreover, the Spearman's correlation coefficient was used to find the relationship between the hemodynamic parameters and the maximum AAo diameter for all patients with BAV together and for each dilation morphotype and fusion phenotype separately. The Spearman's correlation values (S) were discretized in three different groups: good (0.5 < |S| < =0.7), strong (0.7 < |S| < =0.9), and excellent (0.9 < |S| < =1). The statistical analysis was performed using SPSS Statistics (version 25.0 IBM SPSS, Chicago, IL).

A total of 74 patients with BAV (43 men, age 45[37–61] years) and 48 HV (29 men, age 41[30–49] years) were included in the study. Demographic and clinical variables of patients with BAV and HV are shown in Table 1. HVs were matched to patients with BAV in terms of sex, weight, height, BSA, and stroke volume. RL fusion phenotype was present in 66% of patients with BAV. According to the dilation morphotype, 34% patients with BAV (7 BAV root-dilated and 18 BAV non-dilated, or 19 BAV-RL and 6 BAV-RN) were classified as non-ascending aorta dilated (BAV-Non-AAoD) and 66% patients with BAV (19 BAV-RN and 30 BAV-RL) as AAo-dilated (BAV-AAoD).

The three-dimensional maps of the different parameters are shown in Figure 2 for one representative volunteer and two representative patients (one BAV-Non-AAoD and one BAV-AAoD). The phenotype was not included in this figure due to the limitation of selecting a representative morphotype case in each phenotype group. Differences between volunteers and patients with BAV in the AAo and part of the AArch are visible for several parameters: diameter, eccentricity, backward velocity, velocity angle, regurgitation fraction, WSSC, and axial circulation.

From the multivariate analysis test MANOVA, we obtained p-values from Pillai's trace of <0.001 for the HV vs. BAV-All, HV vs. BAV-Non-AAoD, HV vs. BAV-AAoD, HV vs. BAV-RL, and HV vs. BAV-RN. This test showed significant differences between the HV and all BAV groups in all parameters tested. In Figure 3, the comparison between the different BAV groups and HV is shown for each parameter (rows) and aortic region (columns). Most of the analyzed parameters showed significantly higher values in patients with BAV than HV in the AAo and the proximal AArch (regions 1 to 6). In contrast, in the DAo, most of the significant differences resulted from lower values in BAV groups. Eccentricity, backward velocity, velocity angle, regurgitation fraction, WSSC, vorticity, axial vorticity, and axial circulation were consistently greater in patients with BAV than in HV, even in the absence of clinically significant dilation. All parameters (except velocity) showed significant differences in more than one segment for almost all analyzed cases. Velocity, forward velocity, WSS, and WSSA showed lower values in patients with BAV compared to HV.

We also observed similar results for fusion phenotypes (Figures 3B,C) compared to HV. We found that the significant differences in patients with BAV-RN are more concentrated between the ascending aorta and aortic arch, in comparison with BAV-RL, probably, this is influenced by the dilation of the aorta at the proximal part of the aortic arch in patients with BAV-RL, and the descending aorta of these patients show only a few parameters with significant differences. The velocity angle and regurgitation fraction were significantly different for most regions in BAV-RN and BAV-RL. Nevertheless, BAV-RL showed more differences for parameters related to the flow turbulence (parameters from 12 to 17) in the descending aorta with lower values for patients with BAV.

In Supplementary Figure S9, we evaluated the differences between morphotypes and phenotypes groups. BAV-AAoD presented significant differences compared to the BAV-Non-AAoD for velocity, forward velocity, velocity angle, WSS, WSSA, and kinetic energy in the entire aorta. Other parameters such as backward velocity, regurgitation fraction, and vorticity are mostly significant in the ascending aorta. Interestingly, some parameters, which are related to the rotation of the flow, are only significant in the proximal part of the aortic arch, as axial circulation, helicity density, viscous dissipation, and energy loss. Comparing BAV-RL and RN phenotypes groups, we found significant differences in velocity angle, axial vorticity, and axial circulation in more than one region.

Table 2 shows the mean and standard deviation values of each parameter in the AAo (averaged between the regions 1 to 4) for patients with HV and BAV. Patients with BAV presented higher values in most of the parameters in comparison with HV group, whereas velocity, forward velocity, WSS, WSSA, and kinetic energy were lower in patients with BAV. In the AArch, pDAo, and dDAo, the velocity, forward velocity, WSS, WSSA, and the parameters related to the turbulence such as vorticity, axial vorticity, axial circulation, helicity density, viscous dissipation, energy loss, and kinetic energy were lower than in volunteers (Supplementary Tables S1–S3).

To identify the capacity of each parameter to differentiate between BAV and HV, receiver operating characteristic (ROC) curves were calculated. In the AAo, the area under the ROC curves (Supplementary data online Supplementary Table S4) was >0.8 for diameter, eccentricity, backward velocity, velocity angle, regurgitation fraction, WSSC, axial vorticity, and axial circulation (refer to Figure 4). Of note, the velocity angle was the best performing parameter, showing a specificity close to 99% to classify BAV-AAoD from volunteers, which was reduced to 90% in patients with BAV-Non-AAoD. Similar performances were obtained for eccentricity, backward velocity, WSSC, and axial circulation. In the other aortic sections (AArch, pDAo, and dDAo), only one parameter showed an area under the ROC curves >0.8 (Supplementary data online Supplementary Tables S5–S7), the most characteristic being the regurgitation fraction in the AArch sections.

In Supplementary Figure S10, we show the ROC curves for the AAo between morphotypes and phenotypes groups. For patients with BAV-RN vs. BAV-RL, we obtained AUC values lower than 0.63, being the velocity angle the greater value. These results are in concordance with the few significant differences found in Supplementary Figure S9A. Comparing BAV-Non-AAoD vs. BAV-AAoD, excluding diameter and forward velocity, whose values are attributed to the greater area of the segment, the parameter with the highest AUC (0.77) was also the velocity angle.

The MRMR algorithm was applied in each section (AAo, AArch, pDAo, and dDAo), to assess similarities with the ROC curves. In the AAo, the velocity angle, axial circulation, regurgitation fraction, diameter, eccentricity, circumferential WSS, backward velocity, forward velocity, and axial vorticity were the parameters with a greater importance predictor score for most of the cases (HV vs. BAV-All, HV vs. BAV-RL, HV vs. BAV-RN, and HV vs. BAV-AAoD). For HV vs. BAV-Non-AAoD, the velocity angle was the parameter with a greater importance predictor score (Supplementary data online Supplementary Figure S11; Supplementary Table S8). Additionally, in Supplementary data online Supplementary Tables S9–11, we show the predictor important scores for the AArch, pDAo, and dDAo, respectively.

The MRMR was also evaluated between morphotypes and phenotypes only for the AAo (Supplementary Figure S12). The comparison of BAV-RN and BAV-RL shows only three parameters with a higher importance prediction score (vorticity, axial vorticity, and axial circulation), and two of them are in concordance with the parameter found in the Supplementary Figure S9A. For BAV-Non-AAoD and BAV-AAoD, we found that the diameter, backward velocity, WSS, forward velocity, regurgitation fraction, velocity angle, vorticity, and axial vorticity were the parameters with a higher importance prediction score.

In Figure 5, Spearman's correlation coefficient values between the diameter and each hemodynamic parameter in the AAo (the value was averaged between regions 1 to 4) are shown for patients with BAV. Considering all patients with AV together, the aortic diameter presented a strong correlation with forward velocity and velocity angle, the regurgitation fraction, WSS, WSSA, and vorticity presented a good correlation with the diameter. For patients with BAV-RN showed a good correlation for forward velocity and velocity angle. Patients with BAV-RL showed more parameters correlated with the diameter that the other cases, the forward velocity, velocity angle, and WSSA, which are correlated strongly with the diameter, the velocity, backward velocity, regurgitation fraction, WSS, vorticity, axial vorticity, and kinetic energy, have a good correlation with the diameter in this group. For patients with BAV-AAoD, we found a good correlation between the diameter and the forward velocity, WSSA, and vorticity. Finally, patients with BAV-Non-AAoD showed a good correlation for the regurgitation fraction and WSSA.

The main limitation of this study is its cross-sectional nature, which implies that no causal relationship could be inferred. Longitudinal studies are needed to confirm the present results. Another limitation relates to the limited sample size for the group of BAV-RootD, which was included in the non-dilated group, because statistical tests could not be performed on only seven patients. Furthermore, averaging parameters along the circumference of each region can induce sub-estimation of local value. Moreover, the movement of the aorta along the cardiac cycle was not considered in this study since technical limitations in 4D-flow CMR acquisitions (poor contrast and low signal-to-noise ratio) make it difficult to obtain a time-resolved segmentation of the aorta. For that reason, we only analyze the peak systolic cardiac phase. Another limitation of the proposed methodology is that the Laplace approach can be evaluated only for one vessel of interest without branches. Nonetheless, this limitation may also be present in centerline-based methods. The analysis of a large number of parameters can be overwhelming, and for that reason in future contributions, we may generate some artificial intelligence (AI) algorithms that can help to classify groups and categories of these parameters and thus be able to deliver only the parameters that are relevant for a particular pathology, but for this purpose, balanced groups will be needed.