Heart valve disease: investigation by cardiovascular magnetic resonance

Most morphological and functional information is obtained using cine CMR sequences, particularly steady state free-precession (SSFP) sequences with their high contrast between blood pool and surrounding structures (Figure 1). These have largely replaced spoiled gradient echo sequences, though the latter remain useful on occasions for visualising the extent of flow disturbance in selected cases. The ability to image in any plane allows clear views of all four cardiac valves and their inflow/outflow tracts, irrespective of thoracic anatomy or difficult cardiac anatomy. This is particularly useful for right-sided valves which can be challenging to visualise with echo, particularly the pulmonary valve. An additional advantage is that it facilitates direct measurement of valve orifice area for stenosed valves by planimetry rather than calculation [1], and the same technique can occasionally be used for the assessment of regurgitant orifices if required. The anatomical information from CMR cine images can be at least as good as transoesophageal echocardiography, and valvar anatomy and function can often be visualised well with cine images, along with the mechanism of regurgitation, particularly with thin (4-5 mm) slices. There are however several limitations of CMR cine assessment, including a relatively thick imaging slice (typically 5-8 mm) resulting in partial volume effects, and the need to acquire cine images over several cardiac cycles which results in sub-optimal visualisation of small or more chaotically mobile objects such as vegetations. The thin nature of cardiac valves (typically 1-2 mm) makes them particularly prone to partial volume effects due to the slice thickness of CMR images. Care is therefore required in placing image slices perpendicular to the valve plane to minimise these effects and in minimising slice thickness to 4-5 mm, but some aspects of finer valve anatomy may be too difficult to visualise well with CMR. Furthermore, for accurate assessment of the stenotic/regurgitant orifice, positioning the image slice precisely at the valve tips is important and misalignment may result in significant error. Multiple parallel thin image slices in the plane of interest can help to locate the one slice at the optimal position of the valve tips.

The visual assessment of turbulent flow in stenotic or regurgitant flow jets is also feasible with the cine sequences described above, through visualisation of signal voids due to spin dephasing in moving protons [2]. Flow related signal loss seen on SSFP images occurs where voxels span a range of velocities, notably in the shear layers that can surround the more coherent jet core. The location and direction of regurgitant or stenotic jets can be assessed (Figure 2), which can provide valuable information about the valve lesion. In regurgitant jets, the width of the jet origin and/or the vena contracta provide useful information, with milder leaks having narrower jets in general. Lastly, the cine-visualised flow jets can also assist in planning the placement of subsequent velocity-encoded images. Signal voids seen on SSFP imaging are however substantially related to the acceleration of blood rather than the velocity alone, and may underestimate the degree of flow disturbance when assessing the degree of regurgitation. Narrow (mild) jets may be difficult to visualise due to the lack of shear layers at the edge of the jet. Gradient echo sequences are more sensitive than SSFP sequences for evaluating the presence and magnitude of turbulent jets [3], and this sensitivity is increased with lengthening echo time [4, 5]. Assessing the severity of regurgitation with visual assessment of cine images requires care and caution however, as the technique is subject to slice positioning, partial volume effects, the insensitivity of SSFP sequences, and to other sequence parameters. This method can provide an approximate guide to the degree of regurgitation, and distinguishing mild and severe regurgitation is feasible, but finer differentiation of severity is rarely possible.

Accurate measurement of left and right ventricular volumes, function and mass are vital for assessing the impact of valve lesions on the ventricles. Excessive dilation or reduced ventricular function are strong indicators of a poor prognosis [6], and reliable measurement is important. CMR is the most accurate and reproducible technique for assessing both left and right ventricular volumes & mass [7‐9], and newer steady-state free-precession sequences appear to be even more accurate than older gradient echo cine sequences [10, 11]. RV volumes are particularly useful as these are difficult to achieve by other methods, though accurate measurement is more difficult than for LV volumes. The role of left ventricular (LV) mass in valve disease has not been studied as extensively as volume, possibly due to the inaccuracies of measurement by M-mode or 2-dimensional echo [12], and LV mass may become a useful measure in the future, particularly for patients with aortic stenosis. Reproducibility is important for serial assessment of ventricular size, as patients with valve lesions are often monitored for many years if asymptomatic before symptoms and/or ventricular deterioration occur. CMR is highly reproducible [13], and being a 3-dimensional technique, is more sensitive to changes than one or two-dimensional LV diameters [12]. CMR is also less prone than echocardiographic diameters to variations in measurement position, which can occur, despite standard guidelines [14]. The accuracy of CMR is, however, dependent on correct placement of the basal ventricular image slice, and careful contour placement during post-processing is crucial, with correct differentiation of atrial and ventricular chambers, especially for the right ventricle. Significant error can occur if the basal slice is incorrectly included/excluded from ventricular volumes. Post-processing software that includes long axis visualisation of the valves to ensure appropriate slice inclusion significantly aids accuracy. CMR-derived ventricular stroke volumes can also be used to quantify mitral and tricuspid regurgitation, either in conjunction with flow measurement or with volume data alone if the valve regurgitation is isolated [15] - see below, but the issues about accuracy of contour placement still apply.

The ability to quantify flow directly using through-plane phase contrast velocity mapping [16] is a unique advantage of CMR, and does not rely on calculation from complex equations, as echo or invasive catheterisation techniques require. The technique exploits the property of protons moving in a magnetic field gradient, in which they acquire a shift in the phase of their rotational spin as compared with stationary protons, and the magnitude of this phase shift is proportional to their velocity. By producing images from the phase information, velocity can be measured [17], and is visually displayed in greyscale images (Figure 3a). Flow is derived from through-plane velocity maps by integrating the velocity of each pixel and its area over time, typically a single cardiac cycle (Figure 3b). CMR flow measurement shows good accuracy in in-vitro studies and it correlates well with invasive in-vivo measurements [18‐22]. In vivo studies are hampered by the lack of a true 'gold standard' technique for comparison - invasive measures of flow rely on complex calculations and assumptions which may not hold true. The temporal resolution of CMR flow measurement is typically 25-45 msec, which is lower than for continuous wave Doppler velocity measurements in echo, but is good enough for most flow and velocity measurements. Flow measurement is however critically dependent on a homogenous magnetic field, and ensuring the image slice is at the magnet isocentre is important for minimising error. Despite this, phase offset errors due to eddy currents in the magnetic field can still occur, affecting background flow measurements. These are likely to be worse with newer breath-hold sequences due in part to faster gradient switching, and these sequences appear to have poorer accuracy than non-breath-hold sequences [23, 24]. Background flow correction using phantoms can correct the majority of the error [23, 24], and is important for accuracy [25]. The majority of the validation work has been carried out using older non-breath-hold flow sequences, and these are recommended for accurate flow measurement due to the lower background offset error and better validation, particularly if phantom correction is not used. The background offset errors are also worse in very oblique imaging planes [25], and using imaging planes closer to the transverse body plane is helpful [25], where feasible.

Velocities can also be assessed with either 'through-plane' velocity mapping (as used for flow above) or 'in-plane' phase contrast sequences, measuring velocity within the plane of the slice (Figure 4). The in-plane sequences can demonstrate the origin and direction of a jet, and can be useful for visualising the site of stenosis and measuring velocity along the course of a jet, or can assist in planning the subsequent perpendicular or 'through-plane' slice. In-plane sequences may be less accurate however for measuring velocity in a stenotic lesion, particularly peak velocity, for a number of reasons. Due to the image slice thickness (typically 5-7 mm relative to a stenotic jet width of 2-4 mm), they are subject to partial volume effects, in which several velocities occur within a single voxel and an averaged phase shift is measured [26]. The temporal resolution, while reasonable (typically 20-25 msec), is low when compared to fast-changing jet velocities (by comparison, continuous wave Doppler echo temporal resolution can be ~2 msec), and the true peak velocity may be missed. Lastly, the accuracy of CMR velocity measurement in high velocity jets is reduced, particularly with velocities above 3.5-4 m/sec [21, 22]. This is due to signal loss from turbulence [26], and phase shift errors due to fast acceleration and intravoxel dephasing. Utilising sequences with a very short 'echo-time' (~2 msec) can reduce these errors [22] and future applications may use these 'ultra-short' echo time sequences. Narrow, high velocity flow jets (e.g. severe aortic stenosis) are thus especially affected, and the most severely stenotic jets may be too narrow and turbulent for accurate velocity measurement. Through-plane velocity mapping sequences may reduce some of the errors, particularly partial volume effects, as they take advantage of the better within-plane image resolution (typically ~1 mm), to cope with narrow jets. They are thus the preferred method for accurate velocity measurement, though the acceleration & turbulence errors still apply and the temporal resolution is similar. The highest velocity narrow jets may still be too difficult to measure however. Through-plane velocity measurement relies on the plane being placed at the site of maximal velocity (if peak velocity is desired), which is why in-plane velocity mapping to guide placement of the through-plane image can be helpful, unless the site of maximal velocity is predictable. Fortunately, background phase offset errors only result in a small change in the velocity of individual voxels (as opposed to moderate differences when summed for flow quantification), so do not substantially affect velocity measurement.

In addition to flow quantification, recently developed 3-dimensional in-plane CMR flow sequences can measure velocities in 3 dimensions simultaneously [27, 28], allowing the visualisation of complex flow patterns. Future work may examine the utility of this technique for valve lesions and other areas of clinical utility.

The assessment of aortic stenosis is enhanced with CMR through accurate assessment of the anatomy of the valve and aortic root, quantification of LV mass and function to indicate the precise effect on the LV, and measurement of the velocity of the stenotic jet where this is difficult with echo. The excellent visualisation of anatomy provided by CMR allows accurate evaluation of the severity of aortic stenosis [29], usually starting with standard and 'coronal' LV outflow tract views (Figure 5), from which a good qualitative assessment of the valve can be made. Direct planimetry of the aortic valve orifice is the most useful technique for quantifying stenosis severity, achieved by placing an imaging plane through the valve tips in systole (Figure 6). It is important however to ensure the image slice is thin (4-5 mm) and precisely at the valve tips, and acquiring multiple parallel thin slices parallel to the valve orifice may aid in identifying the true orifice [30]. This technique agrees well withmeasured aortic valve area on transoesophageal echocardiography [30‐32] and also with estimated valve area from the continuity equation [31, 32]. Trans-valvar velocity can be measured withvelocity mapping [33, 34], though peak velocity may be less accurate (often underestimated) compared to continuous wave Doppler echo, due to the small width of very high velocity jets (partial volume effects), lower temporal resolution, and artefacts from turbulent jets, as indicated earlier. For these reasons, direct planimetry of the valve orifice area may be a more reliable assessment of aortic stenosis with CMR. It has been suggested that the continuity equation could be used with CMR to estimate valve area [35], with direct measurement of the LVOT area being an advantage over echo. However, this is unnecessary when CMR can directly measure the valve area itself, and the continuity equation relies on multiple measurements and an equation to calculate (rather than directly measure) the valve area, all of which increase the potential for error.

Despite the limitations, CMR measurement of velocity is advantageous in angulated roots where correct echo beam alignment with the stenotic jet is difficult. In addition, many stenotic jets are not parallel to the LV outflow tract, and are also inaccurately assessed with echo, even when outflow tract visualisation is good. In-plane velocity mapping in the outflow tract is useful to identify the location of maximal velocity - this is usually just distal to the valve tips in valvar stenosis. This can be followed with through plane velocity mapping in a plane perpendicular to the direction of flow, positioned at the identified location of maximal velocity (Figure 1). This combination reduces partial volume effects while ensuring the peak velocity is measured, and mean velocity can also be assessed from the through-plane velocity measurements. Ensuring the correct slice position for flow measurement is important for accuracy; and although this image appears similar to that through the valve tips, the position of maximal velocity (the vena contracta) usually lies a few millimetres distal to the valve tips, so the images may not be in identical locations.

Other advantages in aortic stenosis include the ability to differentiate sub-valvar and supra-valvar stenosis, which are easily visualised with CMR cine imaging, and the site of velocity acceleration can be accurately located with in-plane velocity mapping. CMR can also accurately assess the ascending aorta, which may be dilated, particularly with bicuspid aortic valves, and may alter surgical management either at the time of valve replacement or by indicating that root ± valve replacement are required due to excess aortic dilation.

Highly accurate LV mass measurement provides a more precise and sensitive measure of the effect of aortic stenosis on the left ventricle than measuring myocardial wall thickness. LV mass has been a poorly examined parameter in aortic stenosis, likely due to the inaccuracies of echocardiographic M-mode measurement, and CMR-derived LV mass may prove to be a useful tool but needs further examination. Late gadolinium enhancement imaging in patients with aortic stenosis has shown patchy mid-wall enhancement in up to a third of patients with severe stenosis, usually in conjunction with significant LV hypertrophy, and often in the basal lateral wall [36]. This likely reflects focal areas of fibrosis, which have also been shown in autopsy studies [37]. Early reports have shown that this is associated with a worse prognosis [38]. Future CMR techniques may include the assessment of diffuse fibrosis using T1 mapping and other CMR techniques, and this is likely to be an exciting area of development.

The advantages of CMR in aortic regurgitation are quantitation of the regurgitation and of LV volumes and function, particularly for serial measurement. Aortic regurgitation is difficult to quantify with echocardiography, which mostly relies on qualitative and semi-quantitative measures of severity. CMR can accurately quantify the amount of regurgitation using flow mapping, and derived values such as regurgitant fraction (regurgitant volume/forward volume × 100%) can be obtained. Imaging the valve occurs in much the same way as for aortic stenosis, using long axis LVOT views (Figure 7), from which qualitative visualisation of the regurgitation can be performed (Figure 2). Flow can be measured by placing the imaging slice for flow mapping just above the aortic valve, quantifying both forward and regurgitant flow per cardiac cycle [20, 39, 40] (Figure 3). Positioning the imaging slice just above the valve (rather than in the mid-ascending aorta) is important (Figure 7), despite the higher and more turbulent forward velocities encountered here, as underestimation of the regurgitation can occur otherwise [18]. This is due to a number of factors, including movement of the valve towards the LV apex during systole which allows blood to flow into the gap between the imaging plane and the valve. This blood may return to the ventricle during diastole and not flow through the imaging plane, thus being lost to measurement (Figure 8). This tendency is exacerbated by several factors: aortic distension during systole, a dilated aortic root (resulting in greater volume between the imaging plane and the valve), and vigorous longitudinal contraction of the LV (common in significant AR). Placing the plane closer to the valve minimises these errors, though avoiding the very highest turbulence at the valve tips themselves is wise. An ideal flow sequence would incorporate slice tracking to minimise these errors, which tracks the aortic valve and moves the imaging slice accordingly, but this involves complex software programming and has only been performed at a single centre so far [41].

The accuracy of aortic regurgitation quantification using CMR through-plane velocity mapping is excellent when compared to in-vitro studies [42] or in-vivo CMR measurement using the difference between ventricular volumes [20], and it correlates well with angiographic or echocardiographic grades of severity [18, 20, 40]. As it remains the only technique capable of true in-vivo quantification of aortic regurgitation (without calculation), there is no 'Gold standard' for comparison of accuracy. Reproducibility is also good, both for inter-study and intra/inter observer comparisons [39, 40]. The technique is however subject to the same potential problems as all flow techniques, highlighted earlier (under 'flow and velocity quantification'), and care is required to ensure good accuracy of the measurements. In particular, non-breath-hold flow sequences are recommended for their lower background flow offset errors. Quantifying AR with CMR has recently shown a good ability to predict symptom development and the need for valve replacement surgery in the near future [43], with a regurgitant fraction > 33% providing the optimal threshold for identifying patients likely to require surgery within a few years. Given the difficulty in timing valve replacement surgery in patients with severe AR [44], this may become a valuable tool in clinical management, with the potential to identify suitable patients for early surgery. The potential improvement in outcome requires confirmation in a clinical trial however.

AR quantification can also be achieved by a comparison of the differences in LV and RV stroke volume from cine imaging alone [45], but this is less direct and relies on the lack of any other valve regurgitation or shunt. It also assumes accurate contour placement for volumetric assessment, and inaccuracies in measuring any of the four sets of contours (LV and RV in both diastole and systole) can result in significant errors. Careful contour placement is therefore required for this technique, particularly for the difficult RV contours. It is however a useful technique when flow quantification cannot be performed, or as an internal validation of the flow technique. An approximate assessment of the severity of aortic regurgitation can also be obtained by visualisation of the signal void of the regurgitant jet on cine imaging. A narrow jet width at the origin suggests lower degrees of regurgitation, while a wide jet, particularly with a core of high signal from laminar flow, suggests more severe regurgitation. This method is subject to many potential errors however (indicated earlier), and is not recommended for accurate evaluation.

Accurate LV volumes with CMR can aid clinical assessment of the impact of AR, and the high reproducibility is particularly useful for serial assessment, which is important for the management of a condition that has a long asymptomatic phase. CMR-derived LV end-diastolic volumes have also shown some ability to predict the onset of symptoms or other indications for valve surgery [43], and although less strong than quantifying the regurgitation itself, it can provide a useful adjunct in predicting outcome. CMR can also provide a detailed assessment of aortic root anatomy, which can assist in identifying the cause of the regurgitation, and/or whether the root needs replacing at the time of valve replacement surgery.

Taken together, the many CMR techniques useful in AR, including regurgitation quantification, LV volumetric assessment and aortic root anatomy, make CMR the optimal tool for comprehensive assessment.

As for aortic regurgitation, the main advantages of CMR in mitral regurgitation are in quantitative assessment of both the regurgitation, and ventricular volume and function. CMR can also assess leaflet morphology and valve function, particularly utilising the free choice of image planes to characterise the aetiology of the regurgitation in this complex valve. CMR has good agreement with trans-oesophageal echocardiography for assessing mitral valves for repair [46], though the slice thickness can result in partial volume errors more frequently than echocardiography, which has a very narrow beam width. A good method is to place multiple cine images perpendicular to the mitral valve commissure, facilitating assessment of the individual scallops/coaption, and can identify the site of localised prolapse/regurgitation [47]. This technique can be modified by placing three slices specifically across the commissure, perpendicular to the edge of the leaflet at each of the scallops (Figure 9). This results in reduced partial volume effects in the A1/P1 and A3/P3 views particularly, while also being marginally quicker. Despite these good techniques, transoesophageal echocardiography is likely to remain the optimal investigation for leaflet assessment, due to its superior leaflet detail and potential 3-dimensional images. CMR could be used if trans-oesophageal echocardiography wasn't possible, if doubts remained after the echocardiogram, or if other aspects required assessment by CMR and all could be investigated by a single CMR scan. CMR can also be valuable in assessing the regurgitant orifice using thin (4-5 mm) slices parallel to the mitral annulus, carefully placed through the mitral tips in systole [48]. In some patients this can provide direct visualisation and measurement (if required) of the regurgitant orifice area (rather than calculated area from echocardiography), but the complex shape and motion of the valve mean this is not feasible in all patients.

Quantification of mitral regurgitation is usually performed indirectly, mostly by subtracting aortic systolic flow (measured by aortic flow mapping described above) from LV stroke volume [29]. This relies on a combination of two different MR techniques and increases the potential for measurement error, so care is required with the flow sequence and LV contours, as previously highlighted. Mitral regurgitation can also be quantified by comparing the difference in ventricular stroke volumes, as for aortic regurgitation, but the same limitations apply, including the assumption of a single valve lesion and the need for accurate contour placement. The quantification of mitral regurgitation correlates well with echocardiographic and angiographic assessmentand has good reproducibility [49‐51]. The potential clinical utility of quantification is shown by the good ability to predict the future need for surgery in asymptomatic patients [52], similarly to aortic regurgitation, with a regurgitant fraction of 40% providing the optimal threshold for MR. Although this prognostic ability was not quite as strong as for aortic regurgitation, it is significantly greater than echocardiographic semi-quantitation using the proximal isovolumetric surface area (PISA) technique for estimating mitral regurgitant orifice area [53]. Quantitative CMR measures may become a valuable clinical tool to identify suitable patients for early valve surgery, as for aortic regurgitation, though again clinical trials are required to determine if this potential for identifying patients for early surgery translates into improved clinical outcomes. Direct quantification of the regurgitation using CMR is possible using through-plane flow mapping on the atrial side of the mitral valve [54] (Figure 10), but is difficult due to the highly mobile valve and often eccentric, mobile jets of regurgitation. Despite these difficulties, it has reasonable agreement with indirect quantification and can provide an alternative method. Atrial fibrillation is more common in patients with mitral regurgitation, and this can reduce the accuracy of flow measurements [54], but is improved if the heart rate variability is low [54].

Similarly to AR, an approximate assessment of the severity of mitral regurgitation can be obtained by visualisation of the signal void of the regurgitant jet on cine imaging, with a narrow jet width suggestinglower degrees of regurgitation and a wide jet (particularly with a core of high signal) suggesting more severe regurgitation. However, the same limitations apply, and this method is only useful as a rough guide.

Echocardiography, particularly trans-oesophageal echocardiography, remains the first line technique for assessing mitral stenosis due to its excellent visualisation of the mitral leaflets. CMR can be helpful in selected cases however, with good visualisation of the restricted mitral leaflets, particularly on the LVOT view. Direct measurement of the orifice area can be performed in much the same way as for aortic stenosis, by placing an imaging plane at the mitral valve tips during diastole (Figure 11). The technique has good agreement with echocardiography [55], but care needs to be taken in positioning the plane at the tips in order to obtain an accurate valve area, and multiple parallel thin slices may be helpful. Diastolic flow and velocity can also be measured in this image plane, and pressure half time calculated as for echocardiography [56], though the frequency of atrial fibrillation in severe mitral stenosis reduces the accuracy of the flow measurements [54].

The pulmonary valve and right ventricular outflow tract can be difficult to assess with echo, due to several factors. The location of the valve and outflow tract immediately behind the sternum makes it difficult to position the echo probe adequately to visualise the area. The qualitative echo assessment of pulmonary regurgitation is also less robust than for aortic regurgitation, and grading regurgitation severity can be difficult. Thirdly, the right ventricle can be difficult to assess, particularly for volumetric assessment, due to its unusual shape. CMR is therefore particularly valuable for assessing the pulmonary valve, with its combination of free choice of imaging planes, velocity & flow assessment, and accurate assessment of RV anatomy and volumes. Despite the very thin nature of the normal pulmonary valve, making it difficult to visualise with CMR, these other advantages are considerable, and CMR should be considered the 'Gold standard' for assessment of the pulmonary valve and RV outflow tract.

Irregular cardiac rhythms degrade image quality, which affects the assessment of ventricular function, though the effect on this is usually small. The accuracy of flow measurement can also be reduced [54] as flow sequences are acquired over several cardiac cycles (typically 10-12), with data acquired in a complex manner, and reconstructed based on the assumption of a regular rhythm. In subjects with an irregular rhythm, the complexity of the data acquisition means that although acquired over several cardiac cycles, flow is not averaged over these, as might be expected. Where the beat-to-beat variability is small (e.g. atrial fibrillation with a controlled rate), these errors are usually not clinically significant [54]. Very irregular rhythms however (e.g. uncontrolled atrial fibrillation, multiple ventricular ectopics) can present a challenge, particularly to acquiring accurate flow data. Intelligent planning of the timing of data acquisition to the ECG can offset some of the problems, but some patients still present a challenge and caution should be exercised in interpreting the flow data in these. Cine imaging may be more reliable as it is more amenable to intelligent ECG gating and is less affected by arrhythmias, particularly during systole. Ventricular volumes vary with differing heart rates due to differential filling, and these physiological changes do of course still occur, and result in slight blurring of the myocardial borders. The result is an approximate averaging of the volumes over the several cardiac cycles of image acquisition, which is usually acceptable. Flow imaging may be improved with the newer ECG gating techniques for arrhythmias, which typically omit cardiac cycles outside a defined range.

One limitation that remains with CMR is the inability to measure directly the pressure inside a vessel or cardiac chamber - a limitation of all imaging modalities. As for echocardiography, CMR can measure velocity across a stenosis and derive a pressure drop from this, but absolute pressure quantification remains elusive. Cardiac catheterisation remains the most accurate method for assessing this. One paper has identified how CMR may indirectly indicate pressure, by examining the complex flow patterns in the pulmonary artery using 3-dimensional flow imaging [78]. The paper suggested that pulmonary pressure (measured invasively) was strongly linked with the type of flow pattern in the main pulmonary artery. The data requires further validation but indicates the novel approaches to assessment that CMR may bring in the future.