Mitral Valve Flow and Myocardial Motion Assessed with Dual-Echo Dual-Velocity Cardiac MRI

Abstract

Purpose

To develop a dual-echo phase-contrast (DEPC) MRI approach with which each echo is acquired by using a different velocity sensitivity within one repetition time (TR) and demonstrate the feasibility of this approach to measure transmitral blood flow (E) and myocardial tissue (E_m) velocities.

Materials and Methods

The flow across tubes of known diameter was measured by using the proposed DEPC method and compared with flowmeter measurements and theoretic predictions. Then, with both the DEPC MRI sequence and the conventional single-echo phase-contrast (SEPC) MRI sequence, E, E_m, and E/E_m were measured in six healthy volunteers (mean age, 49 years ± 13 [standard deviation]) and eight patients (mean age, 54 years ± 15) being evaluated for cardiac disease. Differences between the DEPC and conventional SEPC MRI methods were assessed by percent error, Pearson correlation, and Bland-Altman analyses.

Results

Velocities measured in vitro and in vivo by using the SEPC and DEPC MRI approaches were well correlated (r² > 0.97), with negligible bias (<0.5 cm/sec) and comparable velocity-to-noise ratios. Imaging times were approximately 19% shorter with the DEPC method (TR, 5.7 msec) than with the SEPC method (TR, 2.8 msec ± 4.2) (P < .05).

Conclusion

The proposed DEPC method was sensitive to two velocity regimes within a single TR, resulting in a shorter imaging time compared with the imaging time in conventional SEPC MRI. Preliminary human study results suggest the feasibility of using this approach to estimate E/E_m.

Supplemental material is available for this article.

© RSNA, 2020

Summary

The proposed dual-echo phase-contrast MRI method can be used to rapidly encode myocardial and transmitral blood flow velocities with sufficient velocity sensitivity within one repetition time to permit extraction of the peak mitral inflow velocity–to–peak myocardial velocity ratio, an index of diastolic function.

Key Points

• ■ The authors propose, implement, and validate a dual-echo phase-contrast (DEPC) MRI method in which each of the two echoes can be sensitized to velocities that can vary by an order of magnitude within a single repetition time.

• ■ Flow rates measured in vitro by using the proposed DEPC method were in agreement with independent measurements of flow obtained by using a flowmeter (range, 0.66–5.08 L/min).

• ■ Mean velocities measured with the DEPC method were in agreement with those measured by using conventional single-echo phase-contrast (SEPC) MRI methods and were within 2% of theoretic predictions.

• ■ In 14 human participants (six healthy volunteers, eight patients), ratios of the mitral annulus velocity to the transmitral blood velocity measured by using the SEPC and DEPC MRI methods were correlated (r² = 0.97, y = 0.98x − 0.05) and showed negligible bias at Bland-Altman analysis (P = .40).

Introduction

The recognition that information regarding tissue motion can be encoded in the phase of MRI signal (1) led to the development of phase-contrast (PC) MRI methods. Numerous clinical applications for time-resolved two-dimensional PC MRI techniques have been found and include quantifying blood flow across vessels and cardiac chambers, estimating pressure gradients, and calculating wall shear stress (2,3). Three-dimensional time-resolved PC MRI techniques (the so-called four-dimensional [4D] flow techniques) have made it possible to study complex flow dynamics in organ systems—for example, in evaluation of aortic disease, congenital heart disease, and neurovascular disorders (4,5)—at the cost of longer imaging times (5–25 minutes) (6,7).

With all PC MRI techniques, the user sets the velocity sensitivity of the acquisition by prescribing a limiting velocity value (called the velocity-encoding [VENC] value) such that spins moving at that velocity acquire an additional phase shift of π compared with stationary spins; tissue motion at velocities higher than the VENC value appears aliased. Setting a VENC value that is high enough to avoid aliasing but low enough to yield a sufficient velocity-to-noise ratio (VNR) (8) can be challenging because neither the maximal velocity nor the dynamic range of velocity within the spatiotemporal extent of the imaging field of view (FOV) is known a priori. Although PC MRI data are routinely acquired with single-velocity sensitivity (ie, with the VENC set slightly above the maximal expected velocity), this approach is suboptimal (9). For example, temporally, aortic blood velocities can be an order of magnitude higher in systole than in diastole. Spatially, the velocities of tissues such as left ventricular (LV) myocardium and blood within the LV cavity can be an order of magnitude or more apart despite their physical proximity to one another. Particularly when flow velocities are measured in vessels across the thoracoabdominal cavity with 4D flow techniques, the large dynamic range of blood flow velocity may limit the use of single-VENC acquisition to evaluating rapidly moving blood in vessels such as the aorta and pulmonary artery, at the expense of a lower VNR when evaluating slowly flowing blood in the venae cavae or the motion of the LV myocardium. To mitigate this problem, several groups have proposed repeating the conventional PC MRI techniques, which we call single-echo PC (SEPC) MRI approaches, with different velocity sensitivities (10,11) over multiple repetition times (TRs) at the cost of longer imaging times. This time cost is particularly great with 4D flow techniques, with which it already takes several minutes to acquire a time-resolved three-dimensional PC volume with single-velocity sensitivity (9,12).

To reduce the total imaging time, we developed a dual-echo PC (DEPC) MRI method with which each echo is acquired by using a user-prescribed velocity sensitivity within a single TR. Phantom experiments were performed to validate this approach. We also tested a potential clinical application of the DEPC method—evaluating the diastolic function of the heart—in both healthy volunteers and patients referred for cardiac MRI.

Materials and Methods

Theory

The data acquired from the study participants were used solely for the current study. VENC gradient waveforms for the DEPC pulse sequence were designed by using the method described by Bernstein et al (13) to minimize the echo time (TE). A DEPC pulse sequence that had dual-velocity sensitivity for tissue motion in the slice select direction (Fig 1) was implemented, and a symmetric “two-sided” VENC strategy was used to minimize TEs and TR in the two segments of the pulse sequence. (See Appendix E1.1 [Velocity-Encoding the DEPC Pulse Sequence] [supplement] for details.) A phase-difference image for each echo was reconstructed by performing the complex subtraction of data acquired from the two segments. The effect on the TR of the DEPC acquisition with the first and second echoes acquired at high and low VENCs, respectively, or from the converse scheme was ascertained with numeric simulations (Appendix E1.2 [supplement]).

Figure 1:

Diagram illustrates the pulse sequence for the proposed dual-echo phase-contrast (DEPC) method involving a symmetric two-sided (solid and broken lines) velocity-encoding (VENC) strategy, with high (red circle) and low (blue circle) VENC for the first and second echoes, respectively. M = frequency-encoding direction, P = phase-encoding direction; RF = radiofrequency; S = slice-encoding direction, TE₁ = echo time for the high VENC measurement, TE₂ = echo time for the low VENC measurement, TR = repetition time.

MRI Measurements

All measurements were obtained by using a commercial MRI unit (Ingenia 3.0T MR System; Philips, Best, the Netherlands). A 16-channel torso phased-array surface coil was used for signal reception, and all data were acquired with simulated (phantom studies) or real (in vivo studies) vector cardiographic triggering. The pulse sequence, developed in house, was applied as a patch to the commercial MRI system.

Flow phantom validation.—A centrifugal pump (Compact⁺ 5000; Eheim, Deizisau, Germany) was used to circulate a blood-mimicking fluid (60% water and 40% glycerol by volume) (14) through two tubes with inner diameters (Ø_ID) of 0.95 and 1.59 cm. These diameters were intended to provide a mean velocity ratio of 2.80 between the two tubes according to the law of conservation of mass (A₁V₁ = A₂V₂, where A₁ and A₂ are the cross-sectional areas and V₁ and V₂ are the velocities of the tubes) (Fig 2, A). The imaging plane was perpendicular to the direction of flow and positioned 30 cm from the inlet and outlet of the flow. An electromagnetic flowmeter (FMS10; Flomotion Systems, Middleport, NY) with a measurement accuracy of 0.4% was placed in line with the flow to provide a non–MRI-based measurement of the flow rate (Fig 2, A), which was controlled by a flow valve. Fluid flow across the tubes was measured by using the DEPC sequence with VENC values set at 180 and 80 cm/sec for the first and second echoes, respectively, and with the following acquisition parameters: FOV, 209 × 209 × 8 mm; acquired voxel size, 1 × 1 × 8 mm³; TR, TE₁, and TE₂, 6.1 msec, 1.8 msec, and 4.5 msec, respectively; flip angle, 15°; number of signals acquired, one; heart rate, 65 beats per minute (at simulated electrocardiography); views per segment, three; temporal resolution, 90 msec per frame; and number of frames acquired, 10.

Figure 2:

Experimental setup for, A, in vitro and, B, in vivo experiments. A, Diagram illustrates a centrifugal flow pump (placed outside the scanner room, along with the reservoir) that pushes blood-mimicking fluid from the reservoir through a larger tube (Ø_ID,1.59 cm) connected to a narrower tube (Ø_ID, 0.95 cm). The flow rate is regulated by using a flow valve (FV) and measured with a flowmeter (FM). The flow of fluid through both tubes is measured across an imaging plane positioned at the isocenter of the imaging unit. A bottle filled with a static fluid (polyalphaolefin) is positioned between the flowing tubes to load the coil and to serve as a reference. (Z points to the direction of the main magnetic field.) B, Conventional four-chamber (I) and vertical long-axis (II) MRI views show a short-axis slice (white bar) prescribed just inferior to the mitral valve plane during systole, used to measure myocardial and mitral inflow velocities with the single-echo phase-contrast (SEPC) and dual-echo phase-contrast (DEPC) MRI techniques. The resulting short-axis image (III) shows the contours used to extract the blood (red outline) and the myocardial velocities (septal [S], lateral [L], and entire myocardium [M] across all cardiac phases). Outer and inner white contours represent the epi- and endocardial boundaries, respectively.

For comparison, conventional SEPC MRI measurements were obtained with the VENC value set at 180 cm/sec and by using a TR (msec)/TE (msec) of 3.3/1.8, and then repeated with the VENC value set at 80 cm/sec and a TR/TE of 3.7/2.1. All other parameters, such as prescribed FOV, acquired voxel size, flip angle, temporal resolution, number of frames, and number of signals acquired, were kept identical between the DEPC and SEPC image acquisitions. Parallel imaging–based acceleration techniques were not used for either method. By adjusting the flow valve, the input flow rates, which ranged from 0.66 L/min to 5.08 L/min as measured by the flowmeter, were measured by using the SEPC and DEPC sequences.

In vivo studies.—Six healthy volunteers (five men; one woman; mean age, 49 years ± 13 [standard deviation]) and eight clinical patients (five men; three women; mean age, 54 years ± 15)—three suspected of having infiltrative cardiomyopathy, three with nonischemic cardiomyopathy, and two with LV hypertrophy—who were referred for cardiac MRI evaluation gave written informed consent to participate in this institutional review board–approved study. A short-axis slice positioned inferior to the mitral valve in systole (planning shown in Fig 2, B, images I and II) was used to measure both myocardial tissue velocity and transmitral blood flow velocity by using the DEPC method with the following acquisition parameters during free breathing: acquired voxel size, 2 × 2 × 8 mm; flip angle, 15°; TR, TE₁, and TE₂, 5.7, 1.6, and 4.4 msec, respectively; high VENC (VENC_H) and low VENC (VENC_L), 150 and 20 cm/sec, respectively; temporal resolution, 18 msec; bandwidth per pixel, 956.5 Hz; views per segment, one; and number of signals acquired, two. For comparison, a conventional SEPC measurement was first obtained with the VENC_H value set at 150 cm/sec and a TR/TE of 2.8/1.6. Then, the SEPC measurement was repeated with the VENC_L value set at 20 cm/sec and a TR/TE of 4.2/3.0. All other parameters, such as prescribed FOV, acquired voxel size, flip angle, reconstructed temporal resolution, and number of signals acquired, were kept identical for the DEPC and SEPC image acquisitions.

Data analyses.—Magnitude and phase-difference images from the PC MRI acquisitions were exported in the standard Digital Imaging and Communications in Medicine format, and custom software written in MATLAB (MathWorks, Natick, Mass) was used for data analysis, which was performed by two reviewers (A.A., R.M.) with more than 5 years of postprocessing experience. Before any further postprocessing, all reconstructed images were corrected for background and eddy current phase contributions by subtracting the phase determined by using the polynomial surface fit of the nonmoving spins within the imaging FOV with use of the method described by Lankhaar et al (15). For the DEPC sequence, background and eddy current phase correction was applied for each echo, and no additional concomitant gradient corrections were applied. All image sets from the clinical participants were reviewed by a cardiovascular radiologist (B.Y.C.) with more than 14 years of experience interpreting cardiac images.

Phantom data analysis.— A mask created by thresholding (Otsu algorithm) the cumulative sum of all acquired magnitude images (10 frames) was used to segment the regions containing flowing spins (16). After magnitude masking, every pixel location in the VENC_H acquisition with a velocity greater than the VENC_L value was identified as a location with an aliased pixel in the VENC_L acquisition (17). Aliased pixels in the VENC_L acquisition were then replaced with corresponding pixels in the VENC_H data to produce a dual-VENC overlapped velocity field. A VNR map of this field was calculated for the SEPC and DEPC methods by dividing the temporal mean of the velocity of each pixel across the cardiac phases by the standard deviation of each pixel, as described by Ha et al (17). The difference in mean velocities between the SEPC and DEPC acquisitions normalized by the mean velocity measured with SEPC imaging was expressed as the percent error. Time-resolved flow rates for the PC MRI methods were computed from the product of average velocity across the tubes and their cross-sectional area.

Analysis of in vivo data.— From the PC MRI data, mean blood velocities across cardiac phases were measured from circular contours drawn within the LV cavity, and mean myocardial velocities across all cardiac phases were measured from annular contours inscribed within the LV myocardium (Fig 2, B, image III) in a diastolic phase. The prescribed contour was propagated frame by frame, allowing the user to adjust the contour to minimize contamination from blood velocities. The peak mitral inflow velocity–to–peak myocardial velocity ratio (E/E_m) in the septal and lateral myocardial walls, a key parameter in evaluating LV diastolic function, also was calculated.

Statistical Analysis

The means (and standard deviations) of all continuous variables are reported. The percentage difference was used to compare mean velocity ratios obtained from the flow phantom experiment by using the SEPC and DEPC methods. Early diastolic peak velocities acquired in vivo were compared by using Pearson correlations, and Bland-Altman analysis was used to estimate bias. P < .05 was considered to indicate a significant difference. All statistical analyses were performed with MATLAB, version R2019b.

Results

Numeric Simulations

Numeric simulations revealed that encoding the first echo with a lower sensitivity (VENC_H) and encoding the second echo with a higher sensitivity (VENC_L) yields a shorter TR (Appendix E1.2 [supplement]). This result determined the arrangement of velocity sensitivities in the pulse sequence (Fig 1).

Phantom Experiment

Flow rates measured by using a flowmeter were highly correlated with flow rates determined from velocity measurements obtained with the SEPC (y = 1.01x − 0.01, r² = 1.0) and DEPC (y = 1.01x − 0.02, r² = 1.0) methods (Fig 3, A). For MRI-based measurements, the total SEPC imaging time was consistently twice the DEPC imaging time (mean time, 1.89 minutes ± 0.40 [standard deviation]) for all VENC pairs. Figures 3, B and C, show representative through-plane velocity images that were acquired from flow phantom experiments by using the conventional SEPC and proposed DEPC methods, with the VENC_H/VENC_L set at 180/80 cm/sec. Note the aliasing in the smaller tube, where flow velocities were higher than the VENC_L (white arrows in Fig 3, C) (8). The overlapped velocity maps obtained by combining the velocity information from VENC_L and VENC_H acquisitions are shown in Figure 3, D. Quantitatively, the spatiotemporal mean and maximal velocities across the larger and smaller tubes measured with the DEPC and SEPC sequences were within 1.1% of each other. A voxel-by-voxel comparison of absolute velocities obtained by using the two methods also indicated excellent agreement (r² = 1, P < .0001) (Appendix E1.3 [supplement]). The theoretically obtained ratio of the mean velocities across the Ø_ID 0.95-cm and Ø_ID 1.59-cm tubes closely mimicked (with a difference of < 2%) the mean velocity ratios obtained experimentally by using the SEPC and DEPC methods (Table). The Reynolds numbers for the flow in the small and large tubes were 3881 and 2248, respectively, indicating nonlaminar transition flow. Velocity profiles and VNR analyses for the flow phantom experiment yielded comparable results for the SEPC and DEPC methods (Appendix E1.4 [supplement]).

Figure 3:

A, Graph shows the flow rates measured with a flowmeter, which were highly correlated to the flow rates obtained from velocity measurements by using single-echo phase-contrast (SEPC) MRI (y = 1.01x − 0.01 [r² = 1.0, P < .0001]) and dual-echo phase-contrast (DEPC) MRI (y = 1.01x − 0.02 [r² = 1.0, P < .0001]) methods. The error bars represent the standard deviation of measurements and were amplified eight times to increase their visibility. Numbers in parentheses are the velocity-encoding (VENC) values, in centimeters per second, used for the MRI-based measurements. B, C, Diagrams show through-plane VENC images in the flow phantom experiment, which were acquired by using conventional SEPC and proposed DEPC methods, with VENC values of, B, 180 cm/sec and, C, 80 cm/sec. The white dotted lines outline the location of the static phantom. Note the aliasing in the narrow tube (white arrows in C) due to the low VENC (VENC_L) being set lower than the actual flow velocities. VENC_H = high VENC. D, Diagram shows the overlapped velocity map constructed from VENC_H and VENC_L acquisition, which is devoid of aliasing.

Velocity Ratios of Flow Through the Flow Phantom Tubes, Assessed by Using SEPC and DEPC Methods and Compared with the Theoretic Velocity Ratio

In vivo experiment.—All participants successfully underwent both SEPC and DEPC data acquisitions with no adverse events. Figure 4 shows representative magnitude and velocity images acquired by using the conventional SEPC and proposed DEPC methods with VENC_H and VENC_L values of 150 cm/sec and 20 cm/sec, respectively, in a healthy volunteer. When the VENC value is much higher than the myocardial tissue velocity, the myocardial VNR is lower (black arrows, Fig 4, A). In contrast, when the VENC value is much lower than the blood velocity, aliasing occurs in the LV blood pool (white arrows, Fig 4, B). The dual-VENC overlapped velocity maps for the SEPC and DEPC methods are shown in Figure 4, C. Figure 5 shows examples of LV blood pool and septal (Fig 5, A, B) and lateral (Fig 5, C, D) myocardial velocities measured in a healthy volunteer and a clinical patient found to have nonischemic cardiomyopathy after presentation. Peak diastolic tissue velocities measured by using the SEPC and DEPC techniques across all 14 participants were correlated (Fig 6, A; r² = 0.97, y = 0.98x − 0.05), and Bland-Altman analysis revealed negligible bias (Fig 6, B; P = .40). The E/E_m, a commonly reported surrogate for LV filling pressure, was within the normal range (<10) in the healthy volunteers, and 62.5% of the clinical patients had an elevated E/E_m (>12), indicating diastolic dysfunction (Table E1 [supplement]) (18).

Figure 4:

Representative cardiac short-axis magnitude and through-plane velocity images acquired in vivo by using the conventional single-echo phase-contrast (SEPC) and proposed dual-echo phase-contrast (DEPC) methods with velocity-encoding (VENC) values of, A,150 cm/sec and, B, 20 cm/sec. Note the greater noise in regions of the myocardium (black arrows) in A (acquired with high VENC [VENC_H]) and the aliasing within the blood pool (white arrows) in B (acquired with low VENC [VENC_L]). The aliased data visible in B were replaced with corresponding pixel values from the VENC_H acquisition by using the dual-VENC overlapping scheme in C.

Figure 5:

Graphs show left ventricular (LV) blood pool and septal and lateral myocardial velocities measured in a healthy volunteer, A and C, and a 44-year-old female clinical patient, B and D. A lower early diastolic peak velocity of the myocardium can be seen in the mean velocity-versus-time curve for the representative patient, leading to a higher ratio of peak mitral inflow velocity to peak myocardial velocity (E/E_m). All velocity data were obtained from the overlapped velocity map. DEPC = dual-echo phase-contrast MRI, SEPC = single-echo phase-contrast MRI.

Figure 6:

Graphs show that absolute values of peak diastolic tissue velocities measured by using the single-echo phase-contrast (SEPC) and dual-echo phase-contrast (DEPC) MRI techniques across all participants were, A, well correlated (r² = 0.97, y = 0.98x − 0.05, P < .0001), and, B, Bland-Altman plots show low mean bias (P = .40). All velocity data were obtained from the overlapped velocity map. SD = standard deviation. Blue indicates myocardium, red indicates left ventricular blood. Squares represent healthy participants, circles represent clinical participants.

Discussion

The results from flow phantom experiments and in vivo human studies show that the proposed DEPC method yields velocity estimates that are comparable to those obtained with conventional SEPC MRI methods. The DEPC approach makes it possible to acquire data that are sensitive to both fast and slow spins within one TR and thus benefit from several advantages.

First, within a large imaging volume such as the thoracoabdominal cavity tissue, velocities can differ substantially, and using a single VENC acquisition to capture blood flow dynamics has an inherent limitation in that the VNR is reduced when spin velocities are far less than the prescribed VENC parameter of the PC MRI experiment. Several groups have recognized this limitation and proposed acquiring PC MRI data with two or more velocity sensitivities (9,12,17) at the cost of a prolonged acquisition time. Using the DEPC method can reduce the imaging time compared with the imaging time in the SEPC method. The extent of this reduction depends on how SEPC imaging is implemented. For example, if the SEPC sequence for each VENC is performed as a separate imaging examination, as is commonly done in clinical practice (and in the phantom and volunteer experiments in this study), then using the DEPC approach instead can shorten the imaging time by as much as 50%. However, if the VENC_H and VENC_L of the SEPC sequence are interleaved on a TR basis, then the imaging time reduction will be in proportion to the ratio of the TRs between the SEPC and DEPC acquisitions, which will depend on the velocity sensitivities chosen (eg, 18.6% {[7.0 msec − 5.7 msec]/7 msec}), as found in in vivo studies). This imaging time reduction is particularly valuable for 4D flow experiments, in which each phase-encoding step has to be acquired over four TRs (to encode velocities in three cardinal directions with one reference acquisition), and repeating the acquisition with use of another velocity sensitivity would entail acquiring each phase-encoding step over seven or eight TRs, resulting in lower temporal resolution, longer imaging times, or both.

Second, although we tested the DEPC method with dual-velocity sensitivity for only a single direction, extending it to volumetric acquisitions and multidirectional VENC would be straightforward. By facilitating a shortened imaging time required for dual-VENC 4D flow PC MRI acquisitions, the DEPC method could enable the widespread clinical adoption of this technique (7). The DEPC phase-velocity mapping method retains all of the traditional advantages of dual-VENC acquisition, such as the potential to use the VENC_H acquisition to unwrap the aliasing in the VENC_L acquisition and thereby improve the VNR (19,20). Furthermore, this method is compatible with the undersampling strategies used to reduce the total acquisition time, such as k-space and time domain under sampling with sensitivity encoding (21), generalized autocalibrating partially parallel acquisition (22), and compressed sensing (23). Furthermore, despite the low specific absorption rate of the gradient echo–based SEPC and DEPC techniques, the radiofrequency burden of the DEPC technique is half that of the SEPC sequence. Thus, the DEPC method may be a preferable alternative for imaging patients with implanted devices who must undergo imaging in accordance with stringent specific absorption rate guidelines (4).

Third, the flexibility of setting two VENC values may prove to be useful in evaluating the rapid ejection of blood across a stenotic valve during systole without compromising the VNR for the measurement of less rapid flow during diastole. It could also allow the simultaneous measurement of fast-flowing blood in arteries and slow-flowing blood in veins, with sufficient sensitivity for both.

Last, our current algorithm replaces aliased pixels in the VENC_L acquisition with those measured with VENC_H and is suboptimal from a VNR perspective. Instead, a phase-unwrapping algorithm that involves the use of information from the VENC_H acquisition could be used to improve the VNR (19,20).

Our study had certain limitations. We compared the flow rates independently measured by using a flowmeter with the PC measurements, and both sets of measurements were in line with theoretic predictions. We acknowledge that independently measuring actual velocities across the tubes by using computational fluid dynamic models or particle tracing facilitates more robust comparisons. Although the background phase correction used in this study for the DEPC acquisition was sufficient to remove any spurious accumulation of phase for the second echo, a more complex background phase correction might be needed to encode velocities in multiple directions. Furthermore, while it is conceivable that more than two echoes with different velocity sensitivities can be acquired within a TR, the potential of such an implementation to improve the VNR or reduce the imaging time would have to be balanced against the confounding effects, such as a reduced signal-to-noise ratio of the later echoes due to longer TEs, sensitivity to intravoxel dephasing in latter echoes, and eddy current effects.

Another limitation in implementing the described DEPC approach is the lack of black-blood suppression for the VENC_L acquisition (24), which is typically used to estimate myocardial tissue velocities. However, we speculate that the conspicuous visibility of blood signal in the VENC_H data set acquired within milliseconds of the VENC_L data set could be exploited to segment the myocardium from the blood pool without the need for additional black-blood preparation methods. Also, we did not independently evaluate diastolic function (eg, with echocardiography) in the human participants, and such a comparison would be necessary before our technique could be clinically adopted. In addition, while use of the proposed DEPC method could conceivably reduce misregistration artifacts arising from two separate acquisitions inherent in the SEPC acquisition, particularly if they were acquired in sequential fashion, we did not evaluate such effects in our study. Last, the results of this pilot study need to be validated in a larger cohort of patients who are known or suspected to have abnormal diastolic function.

We implemented and validated a DEPC sequence with dual-velocity sensitivity within a single TR. In addition, we demonstrated the potential clinical application of the DEPC approach for rapidly estimating the E/E_m—a surrogate for LV filling pressure. The results of echocardiographic studies have shown that the E/E_m is a powerful prognostic indicator in patients with dilated cardiomyopathy (25), congestive heart failure, hypertensive heart disease, and other cardiac conditions, and the DEPC approach described in this article could yield similar information with use of MRI.

We also showed that tissue velocities that differ by an order of magnitude can be measured within a single TR. Using DEPC decreases the amount of radiofrequency deposition by one half and can shorten the imaging time compared with the imaging time of conventional SEPC approaches. Preliminary evaluations in healthy volunteers and clinical patients show the feasibility of this approach for estimating myocardial and transmitral blood flow velocities simultaneously, which can be used to evaluate LV diastolic function. Further clinical studies are necessary to confirm the clinical utility of this approach.