The early detection of changes in RV systolic function is of potential clinical value, particularly in the management of patients with congenital heart disease  but also increasingly in those with acquired pathology where it has been found to have prognostic implications [12–14]. This requires reproducible and readily implemented techniques for clinical assessment of RV contractile function, enabling changes to be identified with confidence on serial follow up. CMR is widely used for the evaluation of congenital heart disease. However, measurements of RV ejection fraction may be a relatively insensitive method for detecting change of RV systolic function with disease [15, 16].
Our data demonstrates the use of a novel CMR tagging technique for the assessment of RV long axis function. Grid or harmonic tagging techniques that have been applied to the left ventricle do not translate readily to the right ventricle owing to its thin walls and complex geometry . Our technique was quick to acquire, in a single additional breath-hold, and to analyse. It recorded differences of RV performance between groups that were not apparent between the measurements of RV ejection fraction. Mean RV displacement was higher in the control (26 ± 3 mm) than in rTOF (16 ± 4 mm) and ASD with pulmonary hypertension (18 ± 3 mm) groups but lower than in the normotensive ASD group (30 ± 4 mm), P < 0.001.
While RV ejection fraction did not differ significantly between the ASD group and the controls, the tagging technique showed less long axis displacement in the ASD group with PHT and more long axis displacement in the ASD group without PHT. This supports the ability of the technique to distinguish ventricular performance between groups of patients with pressure and/or volume loading of the RV, although the previous cardiac surgery, as in the rTOF group, would also be likely to affect the measurements.
Our method of assessing RV long axis function is in principle analogous to the evaluation of tricuspid annular plane systolic excursion (TAPSE) . TAPSE has been shown previously to correlate well with RV ejection fraction by radionuclide angiography in normal subjects and in patients with ischaemic heart disease . This has not been shown in congenital heart disease as very few studies have been undertaken so far in this population of patients. However, the application of the technique of TAPSE is dependent on acoustic access and the angle of insonation , and is not feasible in all patients, particularly after heart surgery or when the RV is dilated. Comparable regions and angles of insonation for TAPSE analysis may be hard to achieve between different operators, although this has not to our knowledge been formally investigated. CMR does not suffer such limitations and complete four chamber views of the heart can be aligned relative to specified anatomical landmarks. We took the mid-point of the mitral valve, the point of angulation of the RV free wall adjacent to the cardiophrenic angle and the apex of the left ventricle as the three points that defined our four chamber plane.
A previous study has attempted to quantify RV long-axis function with CMR. It relied on a manually located point on the tricuspid annulus whose motion was followed relative to the RV apex . This technique introduces potential sources of variation between observers or studies. Another approach used 3D-modeling to estimate the distance traversed during systole by a point at the junction of the RV-free wall and tricuspid annulus . We attempted to optimize our reproducibility by tag placement and automated analysis. The analysis only requires the user to identify the tag line on the image in the first frame (end-diastole) and this is then automatically tracked throughout the cardiac cycle . There is no need to define additional reference points or even to identify end systole. The automated analysis technique achieved narrow inter-observer 95% limits of agreement when compared with the manual technique (± 1.6 mm versus ± 4.9) - a greater than three-fold improvement in precision. The gains in intra-observer reproducibility were of a similar magnitude. The intra- and inter-observer reproducibility we obtained with our manual approach was comparable to that obtained in previous CMR studies using equivalent manual techniques . Furthermore, the reproducibility of the manual technique was comparable to that reported in the echo literature [20, 21]. The automated analysis took less than 2 minutes to complete and so improved precision substantially with little cost in terms of time.
With a view to serial follow up, the automated approach markedly improved inter-study reproducibility (automated versus manual 95% limits of agreement: ± 2.7 mm versus ± 6.7 mm). This greater reproducibility could serve to reduce sample size requirements if the technique were to be used in clinical trials to assess the impact of therapeutic interventions on right ventricular function. Our data provide the basis for sample size and power calculation for future studies in which RV contractile function may be an end-point.
In keeping with previous studies, we found only a moderate but nonetheless statistically significant correlation between RV displacement measured by the automated CMR technique and RV ejection fraction [18, 19]. This is not unexpected as although the two parameters both assess RV systolic performance, RV ejection fraction is a composite of radial and long axis RV myocardial function, LV function as it affects curvature of the septum, and of RV end diastolic volume. As in our study, Konstam et al found that RV ejection fraction in ASD patients with pulmonary hypertension (pressure and volume overload) is similar to that in controls, but impaired in a group with pulmonary hypertension without an ASD (pressure but not volume overload), with preservation of ejection fraction in the former group .
The incomplete correspondence of RV ejection fraction and measurements of RV displacement may also be accounted for to an extent by measurement errors. While measurements of RV ejection by CMR in healthy individuals have shown good reproducibility, the approach may be less reproducible with disease. Grothues et al found that while the inter-study coefficient of variation for RV ejection fraction was 4.3% for patients with normal ventricles, but rose to 10.4% for patients with congestive cardiac failure and 10.0% for patients with left ventricular hypertrophy . With disease, the endocardial border of the relatively trabeculated right ventricle can be increasingly difficult to delineate giving rise to problems with accurately demarcating end-diastolic and end-systolic volumes . As EF is the ratio of the difference in end-diastolic and end-systolic volumes to end-diastolic volumes, errors in the measurement of either of these parameters can summate, increasing the error in the calculation of EF. Our automated technique for the assessment of RV displacement however only requires one measurement in one dimension in a single acquisition, minimising the problem of measurement error.
Our study assessed RV performance at one time point. We did not perform serial follow up of patients to prospectively assess the ability of the novel tagging technique to identify deteriorations in RV function over time. However, the technique was highly reproducible and the differences identified between the pathological groups were far greater in magnitude than the variation resulting from measurement error. Further prospective studies are required to evaluate the utility of this technique in clinical practice.
It is a limitation that we did not prospectively acquire TAPSE measurements by echocardiography in the same patients. When we attempted to collect echocardiographic data retrospectively, we found it to have been sporadic, of variable quality and inadequate for meaningful for comparison with the prospectively acquired CMR data.
A further limitation of the study is that the temporal and spatial resolutions are limited by the need to acquire the data within the duration of a single breath-hold. However, the multi resolution image registration has been shown  to provide sub-pixel tracking of the tagged plane. Furthermore, the temporal resolution captures the principle motion of the cycle and therefore smooth interpolation of the tracked plane between time points enables analysis on a finer timescale.