Author + information
- Received April 14, 2015
- Revision received July 27, 2015
- Accepted August 20, 2015
- Published online July 1, 2016.
- Dimitrios Maragiannis, MD,
- Matthew S. Jackson, MSc,
- Jose H. Flores-Arredondo, MD,
- Kyle Autry, RT(R),
- Robert C. Schutt, MD,
- Paulino A. Alvarez, MD,
- William A. Zoghbi, MD,
- Dipan J. Shah, MD and
- Stephen H. Little, MD∗ ()
- ↵∗Reprint requests and correspondence:
Dr. Stephen H. Little, Cardiovascular Imaging Section, Department of Cardiology, Houston Methodist DeBakey Heart & Vascular Center, 6550 Fannin Street, SM-677, Houston, Texas 77030.
Objectives The aim of this study was to evaluate cardiac magnetic resonance (CMR) phase-contrast (PC) measures of a bioprosthetic aortic valve velocity time integral (PC-VTI) to derive the effective orifice area (PC-EOA) and to compare these findings with the clinical standard of Doppler echocardiography.
Background Bioprosthetic aortic valve function can be assessed with CMR planimetry of the anatomic orifice area and PC measurement of peak transvalvular systolic velocity. However, bioprosthetic valves can create image artifact and data dropout, which makes planimetry measures a challenge for even experienced CMR readers.
Methods From our institutional database, we identified 38 patients who had undergone 47 paired imaging studies (CMR and Doppler) within 46 days (median 3 days). Transvalvular forward flow volume by CMR was determined by 3 methods: ascending aorta flow, transvalvular flow, and left ventricular stroke volume. PC-EOA was derived as flow divided by PC-VTI, calculated with a semiautomated MATLAB (Mathworks, Natick, Massachusetts) application for integration of the instantaneous peak transvalvular velocity. Doppler EOA was assessed by the continuity method.
Results PC-EOA by all 3 flow approaches demonstrated a strong correlation with Doppler EOA (r = 0.949, 0.947, and 0.874, respectively; all p < 0.001) and revealed good agreement (bias = 0.03, 0.03, and 0.28 cm2, respectively). With Doppler-derived EOA as the reference standard, CMR was able to correctly characterize 24 of 26 valves as normal (EOA >1.2 cm2), 12 of 14 possibly stenotic valves (0.8 < EOA < 1.2 cm2), and 5 of 7 stenotic valves (EOA <0.8 cm2; k = 0.826).
Conclusions We describe a new CMR-based method to derive the EOA for bioprosthetic aortic valves. This method compares favorably to traditional Doppler methods and might be an important additional parameter in the evaluation of prosthetic valves by CMR, particularly when Doppler methods are suboptimal or considered discordant with the clinical presentation.
Current recommendations for evaluation of prosthetic valves with Doppler echocardiography provide a detailed assessment of aortic valve prostheses, including peak velocity, peak and mean transvalvular gradient, effective orifice area (EOA), and Doppler velocity index (1); however the Doppler-based evaluation of prosthetic valve function might be technically difficult because of limited imaging windows or might be discordant with the patient’s symptoms or physical examination findings. Cardiac magnetic resonance (CMR) is an alternative noninvasive imaging modality that can be utilized to assess prosthetic aortic valve function by direct planimetry of the anatomic orifice area (AOA) and for assessment of peak through-plane velocity by use of phase-contrast (PC) techniques. In this study, we compare a PC-based CMR method to derive the EOA of bioprosthetic aortic valves (BAVs) against standard Doppler methods in a population of patients with both normal and abnormal prosthetic valve function.
Our institutional database was reviewed to identify patients with BAVs who had both CMR and Doppler studies performed within 60 days. We identified 43 patients who had undergone paired CMR-echocardiography studies between October 2009 and November 2014. All patients underwent CMR scans and echocardiography studies within 46 days (median 3 days). No patient had a cardiac intervention or any new clinical events in the time interval between imaging studies. The clinical indication for CMR included the following: evaluation of ascending aortic aneurysm, pericardial thickness, possible left ventricle (LV) thrombus, mitral valve function, LV pseudo-aneurysm formation, assessment of myocardium viability, and bioprosthetic valve evaluation. Five patients were excluded from analysis for the following reasons: supravalvular aortic stenosis (n = 1), more than moderate paravalvular aortic regurgitation (n = 3), and severe strut artifact (n = 1). Thus, a total of 38 patients (27 men) with 47 paired imaging studies constituted the study cohort. Six patients had a total of 9 additional paired studies for the serial evaluation of aortic aneurysm, aortic dissection, or lung transplant evaluation. These additional paired CMR-echocardiography studies increased the total study population from 38 to 47 paired imaging studies. The mean timing of these follow-up studies was 1 year after the initial imaging study. BAV sizes identified ranged from 19 to 23 mm. The study was approved by the institutional review board.
Cardiac magnetic resonance
CMR studies were performed with a 1.5-T magnetic resonance imaging scanner (n = 37) (MAGNETOM Avanto, Siemens Medical Solutions, Erlangen, Germany) or a 3.0-T scanner (n = 10) (MAGNETOM Viero, Siemens Medical Solutions). Conventional CMR cine assessment with steady-state free-precession sequences provided both morphological and functional data, including left and right ventricular volumes and ejection fraction. A short-axis image of the valve AOA was acquired using 2 orthogonal planes (standard and coronal LV outflow tract views). Consecutive cine images were obtained at slightly modified slice locations (multiple parallel thin slices) to determine the optimal plane to assess AOA at the prosthetic leaflet tips. AOA was measured by planimetry (Figure 1). Imaging parameters were as follows: 1.5-T magnet: slice thickness 4 mm, flip angle 49°, field of view 380 mm, bandwidth 930 Hz per pixel, number of excitations (NEX) = 1, spatial resolution 1.7 × 1.5 mm, and temporal resolution of 38.4 ms; 3-T magnet: slice thickness 5 mm, flip angle 33°, field of view 210 mm, bandwidth 919 Hz per pixel, NEX = 1, spatial resolution 1.6 × 1.3 mm, and temporal resolution of 28.4 ms. Gradient sequences were used when the region of interest was affected by severe susceptibility or steady-state free-precession sequence–related artifacts.
PC pulse sequences were used to assess transvalvular peak velocity and forward flow volume (FFV) at the level of the highest velocities. No motion-correcting algorithms were applied, and acquisitions were obtained during an expiratory breath hold. In-plane PC pulse sequences were acquired for 2 orthogonal planes to locate the direction of the jet. Starting at a perpendicular plane to that jet at the valve tips and using consecutive through-plane PC velocity mapping with appropriate velocity-encoding adjustments, flow and peak velocity were quantified semiautomatically. FFV at the proximal ascending aorta was also measured at the slice level of the sinotubular junction. An experienced cardiologist blinded to echocardiographic results traced PC flow in every frame using semiautomatic contour detection software (Argus, Siemens Healthcare, Erlangen, Germany). The 1.5-T PC imaging parameters consisted of slice thickness 6 mm, flip angle 30°, field of view 400 mm, bandwidth 606 Hz per pixel, NEX = 1, spatial resolution 2.9 × 2.1 mm, and temporal resolution of 38.0 ms. The 3-T PC imaging parameters consisted of slice thickness 6 mm, flip angle 25°, field of view 400 mm, bandwidth 554 Hz per pixel, NEX = 1, spatial resolution 2.9 × 2.1 mm, and temporal resolution of 40.0 ms.
Software was developed in-house with MATLAB (Mathworks, Natick, Massachusetts) to semiautomatically calculate and analyze the PC-derived transvalvular velocity time integral (VTI) by integrating the instantaneous PC-derived peak velocities (Figure 2). Although the velocity points depend on heart rate and temporal resolution, we commonly used 11 data points during the entire ejection period to construct the instantaneous peak velocity plot. No smoothing algorithm was applied to derive the PC-VTI curves. We derived an integral of these peak velocities to calculate the distance of blood travel. By dividing the FFV by this distance measure, we obtain the EOA of the bioprosthetic valve (Figure 3). PC-EOA was assessed by each of 3 different methods of determining the transvalvular FFV: 1) PC-EOAAA: FFV measured at the level of the proximal ascending aorta; 2) PC-EOAAV: FFV measured at the level of highest velocity; and 3) PC-EOALV: FFV measured as CMR LV end-diastolic volume minus end-systolic volume.
All patients had a transthoracic echocardiogram performed with commercially available ultrasound systems. Echocardiographic images were analyzed offline (Digisonics, Houston, Texas). For each Doppler echocardiogram, an experienced cardiologist completed an assessment of BAV function according to American Society of Echocardiography guidelines (1) and was blinded to the CMR results. The following parameters were measured: peak transvalvular velocity, peak transvalvular gradient, mean transvalvular gradient, ejection time, acceleration time, VTI, Doppler velocity index, and Doppler EOA calculated by use of the continuity equation.
Statistical analysis was performed with R software version 3.1.3 (R Foundation for Statistical Computing, Vienna, Austria). Continuous variables were expressed as mean ± SD for normally distributed data and as median (interquartile range) for the non-normally distributed data. CMR parameters of prosthetic valve function, PC-VTI, PC-EOA, and PC-peak velocity, were compared with Doppler VTI, Doppler EOA, and Doppler peak velocity, respectively. Differences between measurements were tested with the paired Student t test or the Wilcoxon signed rank test for matched pairs. Correlation between observed CMR and Doppler values was evaluated with the Spearman rank order test. Bland-Altman analysis (2) was used to assess agreement between methods. To assess for the potential impact of sampling bias, all correlations were reassessed after exclusion of the 9 paired follow-up imaging studies. To evaluate agreement between the 2 techniques, Doppler EOA and PC-EOA were categorized as stenosis, possible stenosis, or normal on the basis of EOA <0.8 cm2, 0.8 to 1.2 cm2, and EOA >1.2 cm2, respectively. Because these variables are ordinal, an equal-weighted Cohen kappa was used to assess interrater agreement between the 2 methods. Interobserver and intraobserver variability was tested in a subset of 10 studies with a 2-way random single measure and a 1-way random 2-measures intraclass correlation coefficient (ICC) analysis, respectively. Statistical significance was attained with a p value <0.05.
The baseline clinical characteristics of the 38 patients studied are summarized in Table 1. Heart rate and systolic blood pressure were not statistically different between CMR and echocardiography studies (p = 0.58 and p = 0.86, respectively). In the 47 paired imaging studies performed, bioprosthetic valve function was characterized as normal in 17 (36%), purely stenotic in 9 (19%), purely regurgitant by variable degrees in 14 (30%; 5 severe, 1 moderate, and 8 mild), and combined stenotic plus regurgitant in 7 (15%; stenotic with mild aortic regurgitation). Five patients had arrhythmia (4 with atrial fibrillation and 1 with a ventricular pacemaker).
Comparison of peak velocity, VTI, and EOA
PC peak velocity correlated well with Doppler-derived values (r = 0.923; p < 0.001), with a small bias (−8.25 cm/s) and 95% confidence interval of −86.45 to 69.95 cm/s (Table 2). Offline VTI calculation demonstrated that median Doppler-derived VTI and PC-derived VTI were 67.6 cm (interquartile range: 39.1 to 90.0 cm) and 63.5 cm (interquartile range: 43.6 to 84.2 cm), respectively (Table 2). PC-derived VTI demonstrated a strong and statistically significant correlation with Doppler-derived VTI (r = 0.918, p < 0.001); however, PC-VTI was consistently slightly smaller (bias of −1.18 cm), with a 95% confidence interval of −21.85 to 19.50 cm.
Doppler-EOA and PC-EOA (calculated using each of 3 different measures of FFV) were calculated, and the results are presented in Table 2. PC-EOAAA and PC-EOAAV both revealed a strong correlation with Doppler-derived EOA (r = 0.949 and r = 0.947; p < 0.001), whereas PC-EOALV correlated well with Doppler EOA (r = 0.874; p < 0.001) (Figure 4). Both PC-EOAAA and PC-EOAAV revealed a slight but clinically insignificant overestimation of EOA compared with the Doppler-derived method. PC-EOALV also showed good agreement with the Doppler EOA technique; however, the bias and limits of agreement were broader than with the previous 2 methods (Table 3). Exclusion of the 9 paired repeated imaging studies did not affect the strength of any correlation assessed (r = 0.953, r = 0.955, r = 0.86, respectively; p < 0.001 for all).
Comparison of anatomic and effective valve area
Assessment of the CMR AOA by direct planimetry was possible in all patients; however, tracing was more challenging in cases of severe leaflet calcification or substantial strut artifact. The AOA was 1.6 ± 0.6 cm2 and correlated well with the Doppler EOA (r = 0.927, p < 0.001), with good agreement between methods (Table 3). As expected, CMR AOA was larger than Doppler EOA (bias = 0.27 cm2), and the limits of agreement were −0.13 to 0.68 cm2.
CMR using the PC-EOAAA method was able to successfully identify 24 of 26 valves with normal EOA (EOA > 1.2 cm2), 12 of 14 possible stenotic valves (0.8 < EOA < 1.2 cm2), and 5 of 7 stenotic valves (EOA <0.8 cm2; k = 0.826). None of the normal valves were reclassified as stenotic or vice versa.
Peak transvalvular velocity was calculated automatically from the contour detection software, and those velocities were plotted during systole with an in-house MATLAB software application. PC-VTI was measured by 2 blinded observers, which yielded a low interobserver variability (ICC = 0.97) and a clinically insignificant mean difference of 1.6 cm (2.4% difference). PC-VTI was also highly reproducible with low intraobserver variability (ICC = 0.99) with a minimal mean difference of 0.2 cm (0.3% difference).
In this study, we have introduced a method to calculate the EOA of normal and dysfunctional BAVs based on PC imaging methods and demonstrated that the PC-derived EOA correlates well with Doppler methods across a range of both normal and dysfunctional BAVs. In addition, we demonstrated that the methods described might have utility for the assessment of bioprosthetic valves with normal function, as well as for the most common modes of valve dysfunction, including stenosis, regurgitation, or combined modes of dysfunction. Likewise, good agreement was found for CMR compared with Doppler standards to identify stenotic, possible stenotic, or normal prosthetic valves.
Quantification of PC-VTI and PC-EOA
In this study, the mean difference between Doppler and PC measurement of peak transvalvular velocity was only 2.7%. Similar results have been reported by Kilner et al. (3), who compared Doppler and PC methods for the determination of peak velocity in a patient group with stenosis of native aortic or mitral valves. Similar accuracy has also been demonstrated in vitro by Fontaine et al. (4), who compared PC and laser Doppler anemometry velocity measurements across prosthetic valves. In our study, we demonstrated the feasibility of using PC velocity measurements to calculate PC-VTI. We demonstrated a strong correlation and high agreement with Doppler-derived VTI measurements. The reproducibility of PC measurements was excellent, which is attributable to the combination of the use of automated software to measure peak velocity and the semiautomatic determination of the PC-VTI.
On the basis of the principle of mass conservation, we calculated the PC-EOA as the FFV divided by the derived PC-VTI. Previous studies in patients with aortic stenosis (5–9) have established the feasibility of a PC-VTI method to evaluate the severity of aortic stenosis. In these studies, the PC-EOA of the native aortic valve compared well with Doppler-derived aortic EOA. In our study, PC-EOA was calculated using 3 different estimates of the systolic forward flow through the valve. All 3 methods correlated well with the Doppler-derived EOA; as expected, the methods that used the flow at the level of the ascending aorta or at the highest peak velocity slice location were superior to the method using the LV stroke volume because they provided a direct measurement of the transvalvular aortic flow.
This is the first study to describe the use of PC velocity information to derive the effective valve area of BAVs. The method is simple and relies on data already acquired during routine clinical imaging of bioprostheses. The ability to obtain a reliable PC-EOA may be useful in clinical practice, because EOA is a parameter used in current guidelines to characterize performance of prosthetic valves and the degree of prosthetic valve dysfunction (1). Although Doppler-derived EOA is the current practice to evaluate aortic bioprostheses, PC-derived EOA might offer an alternative method to assess bioprosthetic valve function when Doppler methods are limited technically. In addition, CMR methods in general might provide a more comprehensive assessment of intrinsic prosthetic valve performance by assessment of both the anatomic and hemodynamic (effective) parameters of the valve. An imaging method that is capable of reporting both the AOA and EOA could be an important clinical tool, because some clinicians might not appreciate that the valve area reported by different imaging modalities may refer to either an anatomic area (e.g., direct planimetry by transesophageal echocardiography or computed tomography) or an effective flow area as defined by Doppler echocardiographic methods. This combined approach to the functional evaluation of prosthetic valves is a unique ability of CMR.
Effective versus anatomic valve area
In current clinical practice, CMR uses the planimetry method for evaluation of aortic bioprostheses in combination with the measurement of peak transvalvular velocities. In our cohort, AOA correlated well with Doppler EOA, which is similar to previous reports. Von Knobelsdorff-Brenkenhoff et al. (10) demonstrated a good correlation (r = 0.82; p < 0.001) between Doppler EOA and the AOA measured by CMR planimetry in a study of 65 patients with predominantly normal (95%) aortic bioprostheses. In our study, a much higher number of dysfunctional BAVs were analyzed, including stenotic valves, regurgitant valves, and valves with combined dysfunction.
AOA provides a single-frame measurement that represents the maximal opening at any individual time point, and AOA is traced at the valve tips, whereas EOA establishes the average vena contracta area occupied by flow at the plane of maximal velocities throughout the entire systolic ejection time period, and the position of the vena contracta is usually a few millimeters distal to the valve tips (11). Pouleur et al. (12) showed that AOA by computed tomographic planimetry did not differ significantly from CMR planimetry; however, planimetric measurements were significantly larger and showed a 16.7% difference between AOA and transthoracic echocardiography–derived EOA in patients with aortic stenosis and normal subjects. Of note, EOA and AOA are related by the following equation: EOA = AOA × cc, where cc is the contraction coefficient. This coefficient varies from 0.6 to 1.0 depending on various parameters, such as the dimensions of the LV outflow tract, inflow geometry, and orifice area (13).
In our patient cohort, the PC-EOA method showed better correlation and narrower limits of agreement than the area planimetry method and had improved reproducibility. The PC-EOA method is less user dependent, because most calculations are semiautomatic. Even among experienced readers, planimetry tracing can be challenging because of partial volume effects, signal dropouts due to calcifications, and strut artifacts. For these reasons, the reporting of a PC-EOA may be a more useful measurement than AOA, because results are more reproducible and more directly comparable to Doppler EOA measurement.
In considering whether CMR might be an appropriate imaging strategy for functional assessment of bioprosthetic valves, the other well-established strengths of CMR should be noted. CMR is considered the reference standard for calculation of LV ejection fraction and has been validated as the modality of choice to assess LV hypertrophy regression and LV reverse remodeling after aortic valve replacement (14,15). Additionally, CMR may provide quantification of markers of diffuse myocardial interstitial fibrosis, which enables estimation of the extracellular matrix components throughout the course of the remodeling process (16). These additional diagnostic markers, along with the PC-EOA, might provide valuable information for the comprehensive evaluation of prosthetic valve function.
One of the limitations of our study was the relatively small number of paired imaging studies included in our analysis. However, to the best of our knowledge, we present the first study to analyze the most common modes of BAV dysfunction by CMR. The slight underestimation of peak velocity by CMR could be associated with different technical limitations. CMR temporal resolution ranges between 25 and 50 ms and is inferior to that of continuous-wave Doppler (17). In addition, velocity and flow measurement errors may occur because of CMR technique limitations (18). In particular, background phase offset errors, ascribed mostly to eddy-current–induced fields, are present in most commercial systems. Additionally, flow can also be underestimated if the slice position is not perpendicular to flow. Nonhomogeneity of the magnetic field may predispose PC measurements to phase offset errors (19). Turbulent post-stenotic flow has also been shown to cause inaccuracies in calculations, especially in cases with high-velocity stenotic jets (20,21). However, novel fast acquisition sequences, such as ultrashort echo time, have been shown to improve the accuracy of velocity measurements (22). Arrhythmias are another potential limitation for an accurate calculation of PC velocities and flow; however, real-time PC sequences might provide a better estimate of the EOA in the near future. Finally, motion of the prosthetic valve plane throughout the cardiac cycle may result in some background flow calculation error; however, motion-tracking algorithms have demonstrated efficacy to eliminate the displacement disadvantage (23).
Of note, our study population consisted primarily of clinically challenging cases referred for both CMR and echocardiography studies within a short time frame. As such, they might not represent the typical referrals for echocardiographic evaluation of prosthetic valve function, and selection bias may be present. However, we strongly believe that this is unlikely to affect the calculation of EOA by the continuity equation and the comparison between the 2 methods, because we have included all forms of BAV dysfunction.
We demonstrate that PC-EOA is a feasible, accurate, and reproducible parameter that may be calculated to assess both normal and abnormal BAV function. Future studies addressing the prognostic value of this novel measurement are warranted.
COMPETENCY IN MEDICAL KNOWLEDGE: Phase-contrast effective orifice area can accurately assess both normal and abnormal bioprosthetic aortic valve function by use of conventional, readily available, and reproducible phase-contrast parameters.
COMPETENCY IN PATIENT CARE AND PROCEDURAL SKILLS: Phase-contrast effective orifice area represents a practical additional parameter to evaluate the hemodynamic performance of prosthetic aortic valves, particularly when Doppler methods are suboptimal or considered discordant with the clinical presentation.
TRANSLATIONAL OUTLOOK: Phase-contrast effective orifice area compares favorably to traditional Doppler-derived effective orifice area; however, its role for outcome prediction has not been investigated. Prospective studies addressing the prognostic value of this novel parameter are needed.
Dr. Maragiannis is supported by the Dunn Foundation for Research and Education. Dr. Little has received research support from St. Jude Medical and Medtronic Inc. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.
Michael H. Picard, MD, served as the Guest Editor for this paper.
- Abbreviations and Acronyms
- anatomic orifice area
- bioprosthetic aortic valve
- cardiac magnetic resonance
- effective orifice area
- forward flow volume
- intraclass correlation coefficient
- left ventricle
- number of excitations
- phase contrast
- velocity time integral
- Received April 14, 2015.
- Revision received July 27, 2015.
- Accepted August 20, 2015.
- 2016 American College of Cardiology Foundation
- Zoghbi W.A.,
- Chambers J.B.,
- Dumesnil J.G.,
- et al.
- Kilner P.J.,
- Manzara C.C.,
- Mohiaddin R.H.,
- et al.
- Caruthers S.D.,
- Lin S.J.,
- Brown P.,
- et al.
- Defrance C.,
- Bollache E.,
- Kachenoura N.,
- et al.
- von Knobelsdorff-Brenkenhoff F.,
- Rudolph A.,
- Wassmuth R.,
- et al.
- Yoganathan A.P.,
- Cape E.G.,
- Sung H.W.,
- Williams F.P.,
- Jimoh A.
- Cawley P.J.,
- Maki J.H.,
- Otto C.M.