Copyright © 2006, The European Society of Cardiology
3D Echocardiography: Is CMR better?
Department of Cardiology, Universitaire Ziekenhuizen Leuven, 3000 Leuven, Belgium
Received 30 May 2006; accepted after revision 30 May 2006.
* Tel.: +32 16 34 49 30; fax: +32 16 34 49 20. frank.rademakers{at}uz.kuleuven.ac.be
The question, as put in this way, cannot be answered in a simple manner since the answer depends on many variables, i.e. type of patient, pathology, circumstances, clinical question, etc. In the setting of a regular operating theater or an intensive care unit, CMR is totally irrelevant since the patient cannot be moved to the MR machine; if the patient has received a pacemaker or an ICD, CMR is relatively contraindicated; if the patient is so overweight that he/she does not fit into the magnet, CMR is impossible, although echocardiographic image quality probably will not be superior either.
The most significant difference is not so much the step from echocardiography to CMR as the one from 2D to 3D. This is true for the appreciation of anatomy but even more so for the accuracy of the determination of cardiac volumes. And cardiac volumes throughout the cycle (including the epicardial volume) provide clinically relevant parameters such as end-diastolic volume, end-systolic volume, stroke volume, ejection fraction and LV mass. Numerous studies have now shown that 3D derived volumes are more accurate than 2D volumes or 1D (M-mode) derived volumes when compared to post-mortem or in vivo CMR studies. The reason has also been clearly established: the fewer the geometric assumptions for the volume calculations, the more accurate the result. With the advent of more computer-aided contouring inter- and intra-observer variability have also improved. Moreover, 3D probes improve the reproducibility of transducer position, because it is easier to control for off-apex positioning causing LV foreshortening. This in turn provides a better inter-study reproducibility. While improved accuracy may not seem that important when one is interested only in a global impression of systolic performance, it may become crucial when deciding on surgical indications of valve lesions or therapeutic adjustments in heart failure. The same holds when volumes are used as endpoints in clinical trials: the more reproducible the technique, the smaller the sample size and the cost of the study. So, not so much the modality as the principles of the technique are relevant for the decision on which one is better, i.e. which one is most suited for the clinical question raised in a specific circumstance.
Since CMR has been around somewhat longer as a 3D modality, several problems have been identified and 3D echocardiography could learn from some of these. Some similarities in the limitations of both modalities are also evident. Although further technical developments will change this, both modalities at present (i.e. in clinical routine) use averaging over several cardiac cycles and are dependent on adequate breath holding by the patient. Stability of probe position during the acquisition for echo finds its counterpart in patient motion during acquisition for CMR. In-plane resolution is much better than through-plane resolution for CMR as is in-line versus lateral resolution for echocardiography. Several strategies have been developed in CMR to correct or to compensate: faster acquisition, respiratory triggering, multiple plane acquisitions with merging of data (taking advantage of the better in-plane resolution of perpendicular planes in an interactive interface), true 3D acquisitions with homogeneous resolution (but with longer acquisition duration), etc. Although most of these may not be (easily) transferred to echocardiography, realizing the limitations is already important to initiate strategies to compensate.
Some other pitfalls of 3D volume measurements should also be mentioned. Correct determination of valve opening and closure is important for accurate volume determination. This is even more important if frame rate is low, as is the case for 3D echocardiography at present. Missing the correct end-systolic frame can cause underestimation of ejection fraction; the smallest area in one short- or long-axis plane is not necessarily the time of the smallest end-systolic volume as this may vary considerably due to regional inhomogeneity; so capturing valve motions and keeping frame rate as high as possible are the best guarantee for optimal timing. The AV valve plane itself is another cause of (remaining) errors in volume determination: ventricular volume is defined as the blood volume confined by the myocardium and the inlet and outlet valves. Especially the AV valve is the structure with the highest compliance of the entire heart, i.e. the AV valve will partially give way under the increased ventricular pressure during systole. Just drawing a line or plane at the level of the AV valve insertions thus incorporates a variable amount of "atrial" blood. Again this will not change an ejection fraction form normal to abnormal but due to the variations in blood pressure during subsequent imaging exams, it contributes to the natural variability in volumes between measurements and to the difficulty to confidently interpret small changes as significant. Measuring blood pressure during the exams for comparison is thus important, and future reconstruction algorithms may be able to use the actual 3D valve contour rather than a simplified AV ring plane.
Improved image quality with CMR, moving from older sequences to steady-state-free-precessing (SSFP), has actually brought a significant problem to the surface which might also be encountered by echocardiography in the future: ventricular trabeculation. Both the left and right ventricles are heavily trabeculated, sparing only small parts of the outflow tract. These trabeculations were not seen on older CMR images, providing a smooth surface for endocardial border delineation. SSFP images, however, show these trabeculations very nicely but make it much more difficult to delineate the endocardial surface: automated segmentation algorithms are incapable at present to follow these small indentations; manually it is nearly impossible and very time consuming to do so; this leaves the choice to either include or exclude these trabeculations or draw a line halfway through, which always introduces a factor of inaccuracy, reduced reproducibility and under- or overestimation of the true ventricular (end-diastolic) volume. Another factor, which was clearly illustrated on these high-resolution cine images, is that these trabeculations merge at end systole, squeezing the blood out from between them and compacting the trabeculations which become indistinguishable from the regular compact myocardial wall. Excluding the trabeculations (at least in part) at end diastole and including them at end systole causes an overestimation of wall thickening and stroke volume. Since these smaller trabeculations are not seen on most echocardiographic images, certainly the 3D ones, one is not aware of this problem in echocardiography, but it remains a factor in the inaccuracy of image segmentation and will need to be addressed in the future when image quality further improves.
Ventricular shape is an important factor in regional wall stress and therefore regional deformation and underlies, at least in part, the observed inhomogeneity of regional deformation. Acquisition of full 3D volumes allows the determination of local radii of curvature and wall thickness, which are important parameters of local wall stress. Improvements in segmentation and computer finite element analysis will allow the calculation of these parameters in clinical practice and will improve the interpretation of regional functional abnormalities as being caused by increased wall stress versus decreased contractility.
In all of the comparisons between imaging modalities one should never forget about image quality per se. Image quality is and remains the most important parameter to obtain accurate, reproducible quantitative data, irrespective of the modality used. And all modalities have circumstances and situations where image quality is suboptimal. When, for one reason or the other, this is the case for a given modality, the patient may be better off to turn to an alternative imaging technique rather than to base decisions on suboptimal data.
In conclusion, going from 2D to 3D is more important for the improvement of image interpretation and quantification of certain parameters than the modality which is used and rather than considering which is better, echocardiography and CMR should learn from each other's strengths and weaknesses. In clinical practice echocardiography is the technique of first choice for a lot of indications, but in case of suboptimal image quality one should not hesitate to move to CMR. Further developments in image processing and analysis should focus on finite element models which can be applied to different imaging modalities and which can provide the common platform for image fusion and comparative interpretation.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  J. Grapsa, D. P. O'Regan, H. Pavlopoulos, G. Durighel, D. Dawson, and P. Nihoyannopoulos Right ventricular remodelling in pulmonary arterial hypertension with three-dimensional echocardiography: comparison with cardiac magnetic resonance imaging Eur J Echocardiogr, November 24, 2009; (2009) jep169v1. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||