Although 2-D echocardiography (2DE) was a major step forwards for the non-invasive assessment of cardiac structure and function, the diagnosis of complex disorders remained a difficult mental conceptualisation process. In addition, the measurement of left ventricular (LV) volume and function, the most common referral reason for echocardiography, required important geometric assumptions about its shape, causing inaccuracy and significant variability. Also, quantitative assessment of the right ventricle with its peculiar shape continued to be elusive. Consequently, imaging methods to display cardiac structures in relationship to each other in the three spatial dimensions were investigated soon after the introduction of 2DE in the 1970s.
The initial approach to 3-D echocardiography (3DE) was the offline reconstruction of a volumetric data set using an external positional locator (acoustic or magnetic) linked to the transducer for recording the spatial co-ordinates of each individual image. This method allowed freehand transthoracic scanning from one or multiple acoustic windows. Gated end-respiratory, end-diastolic and end-diastolic LV contours were obtained and manually traced to produce a wireframe or surface-rendered reconstruction of the LV cavity. The whole procedure was tedious and time-consuming, the images static and there was no tissue information. However, the LV volume and mass quantification from these LV casts proved to be more accurate and reproducible than from 2DE.1
The next-generation 3DE was based on an internal coordinate system in which the transducer can rotate around its central axis to stepwise capture a continuum of 2DE images in a conical volume from a fixed single transducer position. To realise dynamic imaging, a steering logic algorithm, which considers the heart cycle and respiration phase, controls the stepwise image acquisition. A workstation is used to input the images for Cartesian co-ordinate conversion and the reconstruction of a volumetric data set. Rendering algorithms provide cavity and tissue information with the aid of computer interpolation. This method is still used mainly in combination with transoesophageal echocardiography.2
Techniques to shorten the acquisition time have been proposed. Fast continuous rather than stepwise rotational scanning allows scanning of the LV from an apical transducer position and rapidly provides the basic 2-D images for processing and 3-D reconstruction.3
Transthoracic reconstruction techniques for 3-D imaging have not been widely accepted not only because the image acquisition and the offline reconstruction are time consuming but because the image quality is often too marginal for making diagnostic decisions and quantitative analysis. However, the transoesophageal approach has made major contributions by demonstrating the advantages of using a 3-D data set for analysing complex pathomorphology by sectioning structures at various levels and angles, by the en face visualisation of pathology from any perspective. It has been extensively documented with this 3DE technique that the quantification of LV and right ventricular (RV) volumes and function is more accurate and more reproducible than with 2DE by overcoming the problem of cavity foreshortening and avoiding the need for geometric assumptions.
A quantum leap forwards in 3DE was the development of matrix-array transducers that with more powerful microelectronics, allowed the real-time (RT3DE) acquisition and rendering of volumetric data sets. Electrocardiogram (ECG) and respiratory cycle gating are now avoided as well, as motion artefacts result from misregistration of images due to transducer or patient movement during the long acquisition time for the reconstructive techniques.
In an earlier RT3DE system (Volumetrics Medical Imaging) developed at Duke University, a sparse matrix phased-array transducer of 512 elements to scan 60˚ x 60˚ pyramidal volume was used. Parallel processing was applied to permit the reception of 16 lines for each transmitted signal, which allows an imaging rate of 17 pyramidal volumes/sec with a depth of 16cm. This system was mainly used for the simultaneous display of two perpendicular long-axis views or three views orthogonal to the long-axis in selected orientations (C-scans). However, the image quality was rather poor, which has limited its widespread application.
The second-generation RT3DE system (Philips Medical Systems) is based on a complex matrix-array that contains around 3,000 miniature elements. Each of these elements is electronically controlled by a microbeam former incorporated in the transducer head to achieve multidirectional beam steering through a pyramidal volume of 30˚ x 60˚. To capture a data set large enough to encompass the whole (dilated) LV from the apical transducer position, four consecutive ECG-triggered realtime volumes of 20˚ x 93˚ are acquired during held respiration and electronically 'stitched' together in a composite pyramidal data set of 90˚ x 90˚.
Powerful microelectronics process multiple data streams in the transducer and the cardiac structures are reconstructed and rendered in the ultrasound system memory. This process is extremely fast and allows the realtime display of the rendered images (Live3D). Sectioning and cropping away part of the pyramidal dataset allows online assessment of the relationship, orientation and motion of any structure from different perspectives. A zoom mode permits focusing in on a specific structure with higher resolution. An examination protocol for a systematic examination and analysis has been proposed.5
This online display has advantages for the qualitative assessment of valve disease. These can be inspected from any perspective (so-called surgical or en face views) providing direct information on their pathomorphology. In congenital heart disease, defects are visualised and their exact location, size, shape and spatial relationships better appreciated than from 2DE. In recent systems, 3-D colour Doppler reconstruction of regurgitant jets has been integrated and allows the combined display of colour blood flow and tissue data, which significantly increase haemodynamic information, both in valve and congenital heart disease.
The online 3-D display offers also significant advantages for guiding intracardiac catheter-based interventions (defect closure, electrophysiologic studies and ablation procedures, biopsy).
For quantitative analysis the data set is automatically sliced into several equidistant 2-D planes after indicating a few anatomical landmarks. A semi-automated blood-endocardial interface detection algorithm calculates instantaneous cavity contours and a display of their changes during the cardiac contraction. A surface-rendered cavity cast of the LV (or RV) is then constructed, of which the volume is computed without geometric assumptions directly from the voxel counts. This process is automatically repeated every 40ms (imaging rate 25 data sets/second) throughout the cardiac cycle and presented in a volume versus time plot. The calculation of LV (and RV) volume and function is rapid, more accurate and more reproducible than with 2-D and similar to magnetic resonance imaging (MRI), although the variability may be slightly higher as a result of varying image quality.6,7 The availability of the cavity shape allows extraction of additional quantitative information in patients with LV dysfunction (3-D sphericity index).
More user interaction is required to identify the epicardial (and RV septal) contours for the calculation of LV mass.8 However, 3DE has the same accuracy as MRI, but is less expensive and more practical. For the analysis of segmental wall function the surface-rendered cavity casts are automatically subdivided into 17 segments and the centre of gravity of each cast is determined after indicating a few anatomical land-marks. Then the volume of each individual segment relative to the centreline is calculated. This is undertaken for all the datasets captured and a volume versus time curve of the 17 segments is then plotted throughout the cardiac cycle. It is assumed that the segmental volume change reflects segmental wall function. This analysis may offer advantages for detect-ing ischaemic segments during stress echocardiography. A potential advantage of RT3DE is that the standard four 2-D images used for analysis are obtained from the same full-volume data set (four consecutive heartbeats) rather than from multiple sequential transducer positions. This is currently being investigated.
The segmental volume versus time plots allow the measurement of temporal differences in LV wall contraction or dyssynchrony. This technique has great promise for identifying patients who are suitable for resynchronisation therapy and for the immediate evaluation of procedural results and follow-up. Minimal volume (maximal contraction) normally occurs at about the same moment in systole for all segments. In asynchrony, there is dispersion in the timing of reaching the minimal volume as the diseased segments achieve their contraction later. This can be expressed as a systolic dyssynchrony index. Parametric images using a colour scheme representing timing differences in segmental contraction are generated in a 'bull's eye' display, which is a powerful and practical tool for online identification and localisation of LV dyssynchrony.9
Image quality is the most important aspect of any imaging system and is crucial for quantitative analysis of RT3DE. The rapid developments in microelectronics have increasingly contributed to improved image quality and RT3DE continues to evolve rapidly.
A major step forward has recently been made with the introduction of broadband (1-5 MHz) monocrystal transducer technology. These transducers allow high-resolution harmonic imaging with improved cavity delineation. They also offer advantages when left heart contrast agents are needed in patients with less than optimal image quality as the tissue/contrast echo separation is better handled. However, continuous contrast infusion or contrast agents that are more resistant to ultrasound pressure must be used since acquisition of LV full-volume data requires several heart cycles.
The electromechanical efficiency of a monocrystal transducer is 80-100% better than that of the currently used piezoelectric crystals, making them twice as sensitive. This results in better penetration and increased signal to noise ratio. The first clinical results with a transthoracic phased array monocrystal transducer for 2DE and a paediatric matrix-array monocrystal transducer for RT3DE are already available. The image quality and resolution are excellent and competitive with MRI. In the future, a transoesophageal monocrystal matrix transducer for RT3DE will be available as the higher efficiency allows the construction of a small matrix transducer with good imaging performance. Clearly, improved image quality will have a major impact on the diagnostic performance and automated quantitative analysis of 3-D data sets for global and segmental function.
Conclusion
Proven benefits of RT3DE compared with 2DE are the more accurate and reproducible calculation of ventricular volumes, ejection fraction and mass. The results are comparable to MRI. Since, RT3DE is more practical and less expensive it will become the predominant technique for this application in clinical practice. Its use in stress procedures for detecting ischaemia and for diagnosing LV dyssynchrony and testing resynchronisation results remains investigational. However, with the availability of newer sensitive transducers and the next generation of high-speed processors, the spatial and temporal resolution will continually increase and RT3DE is likely to become the most practical and powerful test for detecting ischaemia and LV dyssynergy and monitoring cardiac resynchronisation therapy (CRT) procedures.
The direct visualisation of structure relationships and direct views of dynamic pathomorphology make RT3DE unique for the assessment of valvular and congenital heart disease. The advantages of RT3DE for guiding interventional procedures are increasingly reported. Another important frontier that will be opened by 3DE is virtual reality, the immersive environment created by a powerful computer. Virtual reality allows the cardiologist to interact with these data and heralds a revolution for medical diagnosis, Internet-based distance learning and treatment (e.g. remote robotic interventions).
RT3DE will not replace 2DE but will become an integral part of a complete echo examination, including M-mode (high time resolution for intervals and specific motion patterns), 2-D imaging (accurate dimensional measurements) and quantitative Doppler haemodynamics.