Mitral Valvular Blood Flow & Regurgitation
Accurate and Reproducible Mitral Valvular Blood Flow Measurement and Quantification of Regurgitation
Dutch Heart Foundation, NHS 2000B16
Jos J.M. Westenberg, PhD
Introduction
Accurate measurement of the transvalvular blood flow is invaluable for diagnosing patients with valvular disease. The assessment of mitral valve (MV) regurgitation is currently based on echocardiography. Both twodimensional transthoracic echocardiography (TTE) as well as mono and multiplane transesophageal echocardiography (TEE) with color Doppler are used to provide information needed to plan surgical strategy. Quantification of transvalvular flow velocities based on Doppler measurements has to cope with two sources of error: the alignment of the ultrasound beam to the flow direction and the fixed location of the sample volume. Besides measurement restrictions, the interpretation of echo Doppler measurements to physiological parameters is achieved with the application of geometrical models, which do not apply to all individual subjects. Measurements in multiple planes are necessary, which can yield a summation of errors. Furthermore, imaging the region of interest in TTE can be difficult or even impossible because of acoustic attenuation from structures inside the thorax (lungs, subcutaneous tissues and ribs) and the heart (prosthetic valve construction materials and calcification). TEE is less limited by attenuation of thoracic structures, but this technique is semiinvasive and patients may experience discomfort during this investigation. Magnetic resonance imaging (MRI) is a noninvasive technique, which is readily applied for the determination of global and regional left ventricular anatomy and function. As a threedimensional (3D) imaging technique, volumetric measurements do not rely on application of geometrical models. Velocityencoded cine MRI provides quantitative information on moving spins and can be applied to determine intraventricular blood flow. These promising qualities make MRI a very suitable modality for quantifying transvalvular blood flow. Previous attempts have all encountered the motion of the heart during contraction and relaxation as the main obstacle towards direct acquisition of the transvalvular flow. The mitral annulus moves 1216 mm towards the apex during contraction. Kayser et al. showed that for tricuspid flow quantification, correction for throughplane motion is indispensable (J Magn Reson Imaging 1997, 7, 669673). This also must be considered for mitral flow. Their method corrects for throughplane motion of the right ventricular annulus in a retrospective manner. Since the acquisition plane has a fixed position during the complete cardiac cycle, the location of the data acquisition is not identical to the plane of the MV for the whole cardiac cycle. This implies that flow quantification does not represent the true transvalvular flow. Using velocityencoded MRI, a new acquisition method is introduced to obtain the velocity vector field of the intraventricular blood flow in three directions during the complete cardiac cycle. In this velocity vector field, the MVplane is indicated retrospectively. The velocity values of the vector components going through the MVplane will be reconstructed. For the MVarea, this will result in an accurate MVflow measurement.
Methods
Subjects
Ten patients (8 men,2 women; age 3567 yrs, mean 53 ± 12 yrs) with severe left ventricular dysfunction (left ventricular ejection fraction 36 ± 11%, NYHA class 3.3 ± 0.4) and severe pure MV regurgitation (class 34) were recruited. In movie1, an example of a fourchamber MRI of a patient with MV regurgitation is presented. The MV regurgitation can be depicted from the flow jet from ventricle into atrium during systole. Patients were included consecutively and underwent later a stringent restrictive mitral annuloplasty (2 sizes under) by means of a complete semirigid or pliable ring. Ten volunteers without cardiac valvular disease as confirmed by TTE and TEE (9 men, 1 woman; age 3367 yrs, mean 54 ± 10 yrs) were recruited for comparison purposes. All patients and volunteers gave informed consent and approval from the Medical Ethical Committee of our hospital was obtained on all examinations.
MRI Acquisition
MRI was performed on a 1.5 T scanner (ACSNT15 Gyroscan with the Powertrack 6000 gradient system; Philips Medical Systems, Best, The Netherlands), using the body coil for transmission and a five element phased array cardiaccoil placed on the chest for signal reception. First, scout images and two and fourchamber acquisitions as well as a complete shortaxis acquisition were performed (conform standard cardiac MR protocols, using balancedFFE), needed for planning the velocityencoded MR scans. To obtain the velocity vector field of the intraventricular blood flow, a multislice spoiled gradientecho (phasecontrast) sequence with velocityencoding in all three directions was acquired as follows. A radial stack of six imaging planes was positioned on the left ventricle (LV), with an interplane angulation of 30° and the radial axis of the stack going through the MV annulus and the apex (both found in a short axis view), coinciding with the long axis of the LV (Figure 1). Retrospective cardiac synchronization was used, 30 phase images were reconstructed for one cardiac cycle. Typical examination time was 912 minutes, depending on the subject's heart rate. In each consecutive imaging plane, velocity values were acquired in three directions, with a maximum velocity sensitivity value of 100 cm/s in each direction (movie2). The reconstructed velocity values in three directions represent the velocity vector components. From these velocity vector components, the intraventricular velocity vector field in 3D can be reconstructed (movie3).
Figure 1: Planning a radial stack of 6 LA slices for 3D VE MRI.
MVFlow Quantification
From the 3D velocity vector field of the intraventricular blood flow, the flow through the MV was reconstructed. To achieve this, the MV plane was determined for each of the cardiac phases in the 3D velocity vector field. In each of the six acquisition planes the mitral valve annulus was manually indicated by a straight line positioned on the closed valve leaflets during systole and on the annulus and perpendicular to the left ventricular inflow direction during diastole (movie3). The velocity components perpendicular to each of the six MVlines were projected onto a single twodimensional (2D) plane, representing the MVplane. Triangular interpolation between the sample points on the six lines was used to obtain velocities for the complete 2D MVplane (movie4).
The velocities measured perpendicularly to the reconstructed MVplane need to be corrected for the motion of the myocardium in basal/apical direction in order to obtain the true transvalvular velocity of the blood flow. The velocity of the MV annular plane was obtained from the two and four chamber acquisitions. From the displacement of the MV annulus in these images, the throughplane velocity of the MV was determined. This velocity value was subtracted from the throughplane MVflow velocities measured at the reconstructed plane, resulting in the corrected velocity values with respect to the MVannulus. Finally, the transvalvular volume flow was obtained after manually drawing a contour in the reconstructed velocity image of each phase, containing the transvalvular velocities and integrating these velocities over the area. The MVflow volume was obtained by calculating the Riemann sum of the flow graph (example in Figure 2).
Figure 2.
Validation
For each subject, the MVflow volume determined with the new method described above (from hereon noted as the "3dir MVflow") was validated by comparing it with the flow volume measured with MRI at the ascending aorta, referred to as the AOflow volume. The same MRI acquisition protocol was used as for the MVflow, but now with a single imaging plane positioned perpendicular to the ascending aorta. Only velocity values in throughplane direction, perpendicular to the acquisition plane, were considered. The aortic flow was determined using the QFlow software (Medis, Leiden, The Netherlands).
Furthermore, for each subject, the 3dir MVflow volume was compared with the conventional onedirectional MVflow volume measurement with throughplane motion correction. The same imaging protocol was used, with the single imaging plane positioned at the MV during endsystole. Again, only the velocity values in throughplane direction were considered. The throughplane velocity of the myocardium was assessed in a regionofinterest, drawn inside the myocardium for each of the phases, by measuring the mean velocity inside this regionofinterest. This velocity value was subtracted from the velocity measured at the MV annulus. The MVflow volume measured with this method is referred to as the "1dir MVflow volume".
For ten volunteers and ten patients, the MVflow volume was determined with the 3dir MVflow method, as well as with the 1dir MVflow method, and compared with the AOflow. The regurgitant flow volume in the patients was determined by calculating the flow volume during systole going from the left ventricle into the left atrium. The regurgitant flow fraction, a measure for the severity of regurgitation, was determined by the ratio between the regurgitant flow volume during systole and the inflow volume through the MV during diastole.
The reproducibility of the interpretation of the velocity images, acquired with the new 3dir MVflow method, was tested by inter and intraobserver studies. Two observers analyzed the 3directional MR velocityencoded images of the healthy volunteers to study interobserver variation; one observer analyzed the images twice, with an interanalysis time of more than one week, to study intraobserver variation.
Figure 3.
Figure 4.
Results
The new 3dir MVflow method was first tested on ten healthy volunteers and compared to the conventional 1dir MVflow. The effective forward flow volume measured at the MV with either method, was compared to the flow volume measured with 1directional throughplane velocityencoded MRI at the Aorta (AOflow). The results are presented in Figure 3A. KolmogorovSmirnov tests were performed to prove that the data of each of the parameters were distributed normally (p = 0.84 for 1dir MVflow, p = 0.99 for 3dir MVflow, p = 0.89 for AOflow). The correlation between the MVflow volumes acquired with both methods respectively, and the AOflow volume was tested with Pearson correlation coefficients. The correlation between the 1dir MVflow volume and the AOflow volume is not statistically significant: (r_{p} = 0.15, p = 0.68), whereas the correlation between the 3dir MVflow volume and the AOflow volume is statistically significant: (r_{p} = 0.92, p < 0.01).
Differences between the MVflow volume and the AOflow volume were analyzed following the approach described by Bland and Altman. The results are presented in Figure 3B. The 1dir MVflow volume shows an overestimation compared to the AOflow volume. The mean difference ± standard deviation = 25 ± 22 ml, and this difference is statistically significant (p < 0.01). On the other hand, no systematic difference between the 3dir MVflow volume and the AOflow volume is found. The mean difference Â± standard deviation = 5 ± 7 ml (p = 0.06).
For studying the reproducibility of the results obtained from analysis of images from the new 3dir MVflow method, an intra and interobserver study was performed. The results are presented in Table 1. There is very good correlation between the flow volumes from both analyses, (r_{p} = 0.97, p < 0.01). The differences between both analyses were studied conform Bland and Altman, and are also presented Table 1. There is no statistically significant difference between both analyses (p = 0.61). The mean difference ± standard deviation is 0.9 ± 5.1 ml.
For the interobserver study, also a second observer performed the image analysis, and the results were compared with the results of the first analysis of the first observer (also Table 1). There is an excellent correlation between the flow measurements from both observers, (r_{p} = 0.96, p < 0.01). The differences are compared to the mean 3dir MVflow volume of both analyses. There is no statistically significant difference between the results from both observers (p = 0.49). The mean difference ± standard deviation is 1.3 ± 5.6 ml.
Table 1 
Observer 1 2^{nd} 
↔ 
Observer 1 1^{st} 
↔ 
Observer 2 
KolmogorovSmirnov 
p = 0.98 

p = 0.99 

p = 0.43 
Pearson correlation 

r_{P} = 0.97 

r_{P} = 0.96 (p < 0.01) 

Paired ttest 

p = 0.61 

p = 0.49 

Mean difference 

0.9 ml 

1.3 ml 

Standard deviation 

5.1 ml 

5.6 ml 

Observer 1 performed the analysis twice. The results of both analysis are compared. The results of the first analysis by observer 1 are compared with the results from the analysis of observer 2. To study differences, a paired ttest is performed. Correlation is investigated by the Pearson correlation coefficient.
Next, the new method for measuring the MVflow volume was performed on the ten patients selected for mitral valve repair. The regurgitant flow fractions were between 3% and 30% (mean ± standard deviation = 16 ± 8%) in these patients. The results of the MVflow measurements are presented in Figure 4A. Again, KolmogorovSmirnov tests were performed and prove that the flow values can be considered as distributed normally (p = 0.98 for 1dir MVflow, p = 0.91 for 3dir MVflow, p = 0.68 for AOflow). The correlation between the 1dir MVflow volume and the AOflow volume is not statistically significant: (r_{p} = 0.08, p = 0.82), whereas the correlation between the 3dir MVflow volume and the AOflow volume is statistically significant: (r_{p} = 0.90, p < 0.01).
The differences between the MVflow volume and the AOflow volume are presented in Figure 4B. The 3dir MVflow volume shows no systematic difference with the AOflow volume (mean difference ± standard deviation =  4 ± 7 ml (p = 0.15)). For the 1dir MVflow, the mean difference Â± standard deviation = 19 ± 28 ml, but this difference is not statistically significant (p = 0.06) due to the large standard deviation.
Conclusion
The new 3dir MVflow method provides small, nonsignificant differences compared with the AOflow volume and good agreement between the flow measured at both locations.This 3dir MVflow method was subjected to an intra and interobserver variation study. A high reproducibility was found, excellent correlation between repeated analyses and the differences between analyses were very small and not statistically significant.
The new 3dir MVflow method enables accurate and reproducible quantification of the true transvalvular MVflow in a patientfriendly and easytouse manner. The threedimensional velocity vector field of the intraventricular blood flow enables quantification of asymmetric regurgitant flow jets, a limitation of TEE which can miss asymmetric jets due to prolapse of the commissurae.
The new method can be applied on most commercially available MRI scanners.
Status
Finished.
Publications
 Westenberg JJM, Danilouchkine MG, Doornbos J, Bax JJ, van der Geest RJ, Labadie G, Lamb HJ, Versteegh MIM, de Roos A, Reiber JHC. Accurate and reproducible mitral valvular blood flow measurement with threedirectional velocityencoded magnetic resonance imaging. J Cardiovasc Magn Reson 2004; 6 :767776.
 Westenberg JJM, Doornbos J, Versteegh MIM, Bax JJ, van der Geest RJ, de Roos A, Dion RAE, Reiber JHC. Accurate quantitation of regurgitant volume with MRI in patients selected for mitral valve repair European Journal of Cardiothoracic Surgery 2005; 27:462–467.
 Westenberg JJM, van der Geest RJ, Lamb HJ, Versteegh MIM, Braun J, Doornbos J, de Roos A, van der Wall EE, Dion RAE, Reiber JHC, Bax JJ. MRI to evaluate left atrial and ventricular reverse remodeling after restrictive mitral annuloplasty in dilated cardiomyopathy. Circulation 2005; 112 [suppl I]:437442.
 Westenberg JJ, Braun J, Van de Veire NR, Klautz RJ, Versteegh MI, Roes SD, van der Geest RJ, de Roos A, van der Wall EE, Reiber JH, Bax JJ, Dion RA. Magnetic resonance imaging assessment of reverse left ventricular remodeling late after restrictive mitral annuloplasty in early stages of dilated cardiomyopathy. J Thorac Cardiovasc Surg. 2008; 135:12471252.
 Westenberg JJ, Roes SD, Ajmone Marsan N, Binnendijk NM, Doornbos J, Bax JJ, Reiber JH, de Roos A, van der Geest RJ. Mitral Valve and Tricuspid Valve Blood Flow: Accurate Quantification with 3D Velocityencoded MR Imaging with Retrospective Valve Tracking. Radiology. 2008; 249:792800.
 Roes SD, Hammer S, van der Geest RJ, Ajmone Marsan N, Bax JJ, Lamb HJ, Reiber JHC, de Roos A, Westenberg JJM. Flow assessment through four heart valves simultaneously using 3dimensional 3directional velocityencoded magnetic resonance imaging with retrospective valve tracking in healthy volunteers and patients with valvular regurgitation. Invest Radiol 2009;44:669675
 Marsan NA, Westenberg JJ, Ypenburg C, Delgado V, van Bommel RJ, Roes SD, Nucifora G, van der Geest RJ, de Roos A, Reiber JC, Schalij MJ, Bax JJ. Quantification of functional mitral regurgitation by realtime 3D echocardiography: comparison with 3D velocityencoded cardiac magnetic resonance. JACC Cardiovasc Imaging. 2009;2(11):12451252.
 Brandts A, Bertini M, van Dijk EJ, Delgado V, Marsan NA, van der Geest RJ, Siebelink HM, de Roos A, Bax JJ, Westenberg JJ. Left ventricular diastolic function assessment from threedimensional threedirectional velocityencoded MRI with retrospective valve tracking. J Magn Reson Imaging. 2011;33(2):312319.
Contact
Jos J.M. Westenberg, Ph.D.
Division of Image Processing
Department of Radiology, 1C2S
Leiden University Medical Center
P.O. Box 9600
2300 RC Leiden
The Netherlands
Tel. +31 (0)71 526 2138
Fax. +31 (0)71 526 6801
email: J.J.M.Westenberg@lumc.nl