Introduction

Technologic advances continue to propel cardiac visualization forward. Advances in miniaturization, computer processing, and algorithms continue to have significant momentum. New acquisition technologies, coupled with the ability to process, display, and quantify enormous amounts of data, have pushed echocardiography forward into becoming a modality that will change cardiac intervention as well.^1–13 Today, three-dimensional transesophageal echocardiography (3DTEE) has generated near-optical cardiac images. An understanding of cardiac mechanical motion in all of its spatial and temporal dimensions is providing valuable general physiologic insight for echocardiography as well as precise diagnoses for patients.

It is expedient to address the technologic advances in terms of the operating sequence in an ultrasound system for echocardiography. This allows a framework consisting of transduction, beamforming, display, and quantification to be used. Echocardiographic instrumentation is unparalleled with respect to its portability, cost, lack of ionizing radiation, and ubiquitous presence.

Transducer Technology

Among imaging modalities, the defining aspect of echocardiography is the transducer. Transduction refers to the ability to convert one form of energy to another. Thus, an ultrasound transducer converts electrical energy into mechanical energy on “transmit” and acts as a microphone on “receive.” All ultrasound transducers perform this operation, but what sets ultrasound advancements apart is the ability to steer in two or three dimensions. The earliest M-mode transducers created an image composed of one spatial dimension and one temporal dimension. The element is composed of specialized material that traditionally uses lead-zirconate-titanate. Newer single-crystal materials that contain homogeneous solid-state domains are more efficient in the transduction process and have higher bandwidth (more upper and lower frequency content). This creates a concomitant increase of echo penetration and resolution. Although the M mode was an advance over the stethoscope, it was limited by its lack of field of view. M mode used a “transmit-listen-wait” duty cycle to determine the distance of targets along an unsteered scanline, and the operator needed to point the transducer to examine different cardiac structures. The development of the phased-array paradigm allowed scan lines to be steered.

Conventional two-dimensional (2D) imaging, commonly used in echocardiography, uses transducers with several elements (Figure 3-1). Specifically, elements are oriented in a single row containing 48 to 128 elements, with each element electrically isolated from the others. Individual wavefronts are generated by firing elements in a certain sequence. Each element, constructively and destructively, adds and subtracts pulses, respectively, to generate a focused wave that has direction. This creates the radially propagated scan line. For example, if the farthest element on the right fires first and a timed sequence propagates along the element to the left, the beam will be steered to the left (Figures 3-2 and 3-3). Each element fires with a delay in phase with respect to a transmit initiation time. To further clarify this point, this one-dimensional array of elements can fire in two dimensions: radially and azimuthally (laterally). This spatiotemporal orientation of elements and their phase-timed firing sequence form the underpinning of any modern phased-array system.

Figure 3-1 Transthoracic imaging transducers for three-dimensional echocardiography in the adult. Three generations of transducers from left to right: ×4, ×3-1 and ×5-1. Miniaturization is the key technological trend in the evolution of matrix array technology. The ×4 has 24 application-specific integrated circuits and the ×5 has one. This reduces power consumption and also allows a more ergonomic handle. Other changes that have evolved include changes to the physical and nose apertures to optimize acoustic coupling in intercostal spaces.

Figure 3-2 Integrated circuit miniaturization in the ×5-1 transducer. The inset shows previous technology using multiple application-specific integrated circuits (left). These perform the micro-beamforming and allow steering of acoustic lines in three-dimensional space both azimuthally and elevationally. Inset at right, These circuits have been condensed to one circuit whose dimensions are designed to match the aperture shown in red on the transducer face.

Figure 3-3 Portion of a three-dimensional beamforming application-specific integrated circuit (ASIC). Shown are electrical contact points that couple the transducer piezoelectric elements to the beamforming component circuitry. The ASIC is fully capable of performing M-mode, two-dimensional, and 3D imaging as well as spectral and color Doppler.

Transducer material is cut, or “diced,” by a diamond-tipped saw to create the checkerboard pattern. Elements are then electrically connected to a system. Early-generation systems were composed of sparsely sampled arrays, that is, arrays whose elements were not all electrically active. These sparse arrays created the first instantaneous 3D images; early clinical research was conducted using this type of transducer. It is advantageous to have every element independently active to control the ultrasound beam with more precision. The spacing, or pitch, of these elements depends on the desired frequency of operation (typically λ/4). Otherwise, undesirable diffraction effects such as grating lobes appear. This means that higher frequency transducers have finer pitches. Increased technologic challenges have emerged in the creation of these element connections. The major advance that allowed a fully sampled matrix array to be fabricated was the ability to develop electrical interconnections to every element.

The key aspects pertinent to 3D echocardiography (3DE) imaging involve imaging moving structures. If a static structure that is not moving in space needs to be visualized, a 2D imaging transducer could be swept if the third spatial dimension could be additionally registered within the coordinate space. By using electromagnetic trackers, early gated 3D methodologies exploited this paradigm. Naturally, the 2D images needed to be gated to the electrocardiogram (ECG). This could lead to error caused by movement or arrhythmias, and these lengthy scans could take tens of minutes, requiring a few hundred heart cycles. Ultimately, high-resolution cardiac imaging requires instantaneous imaging to overcome these limitations. The key difference between a 2D imaging transducer and an instantaneous 3D imaging transducer is the arrangement of elements (Figures 3-4 to 3-7). Although a one-dimensional row is used for 2D imaging, a 2D matrix or checkerboard is used to steer an ultrasound scan line in the azimuthal as well as the elevational direction. A conventional 2D imaging transducer of one row steers energy in the azimuthal plane but unintentionally propagates elevational energy above and below the scanning plane. This checkerboard pattern allows phasic firing of elements to generate a radially propagated scan line that can be steered laterally and in elevation. Thus, the true 3D imaging transducer is born.

Figure 3-4 Internal component of a transthoracic three-dimensional imaging probe. Shown at the bottom is the cable coupling to the housing. The electrical cable connections to the main beamformer in the ultrasound system pass through the rectangular connector (shown in light green) in the middle of the handle. At the top is the beamforming application-specific integrated circuit and matching layer.

Figure 3-5 Three-dimensional transesophageal imaging transducer.

Figure 3-6 Three-dimensional transesophageal echocardiography imaging probe (left) compared with a multiplane probe (right).

Figure 3-7 Three-dimensional transesophageal echocardiography (TEE) imaging probe tip with external housing removed. Shown is the matrix array in both azimuthal and elevational dimensions. Unlike a conventional multiplane TEE probe, there are no moving parts within the housing. Moreover, special shielding is built in to prevent electrical interference from external sources such as electrocautery.

Micro-beamforming is the process of using coarse and fine steering. This is implemented by putting fine-delay circuitry into special, application-specific integrated circuits (ASICs). The first commercial, fully sampled matrix array transducer used this methodology by placing 24 to 26 ASICs into the transducer handle. Approximately 3000 elements were electrically connected to these ASICs. Fine steering was performed using subsections of the element matrix known as patches. Coarse steering is performed within the system and through a conventional cable. Specially engineered ASICs allow an individual element to be electrically active but simultaneously keep the size of the transducer cable small, since a significant portion of beamforming has already taken place in the handle. Early transducers were specialized “3D-only probes,” but it is now possible to perform all the transducer functions such as imaging, color flow, and spectral Doppler within the same transducer. Moreover, the transducer aperture should be large enough to allow sufficient lateral and elevational resolution but small enough to “fit” into the intercostals space. One of the most difficult aspects of transducer engineering is what is known as thermal management. The electronics generate heat, potentially more so at high mechanical indexes (e.g., higher waveform amplitudes). These issues need to be resolved if 3DE was to move to the operating room.

3DE depends on micro-beamforming miniaturization. By reducing the electronic substrate required to beamform onto a single chip, the transducer chip is miniaturized sufficiently to pass into the human esophagus. It also significantly reduces the power requirement and hence the amount of heat generated by live circuitry (Figures 3-8 to 3-10).

Figure 3-8 Three-dimensional transesophageal echocardiography transducer with probe handle components displayed. Shown are steering cables for the bending neck. The bending neck, which is proximal to the probe tip, allows anteroflexion and retroflexion within the esophagus. Four electrical connectors allow post–micro-beamformed acoustic information to pass within the cable.

Figure 3-9 Three-dimensional (3D) steering acoustic lines in 3D space. By steering transmit lines, a fully functional matrix array can interrogate anatomy in 3D.

Figure 3-10 Three different volume scanning paradigms. At left is a volume used for live scanning. The center shows live scanning but in a smaller zoom fashion. At right is a larger volume with multiple gated slabs put together.

Beamforming

3DE beamforming consists of steering and focusing of ultrasound energy both as transmitted and received scanlines. This creates useful signals that can be displayed or quantified. It is both advanced science and art. Significant advancements that maximize frame rate, scanning volume size, and resolution continue to occur. Resolution is defined as the ability to distinguish two point targets as distinct. The limiting item in current 3DE systems is the speed of sound, not computing power. Ultrasound image quality improves by firing more transmit lines with more closely associated line spacing. This slows the frame rate since there are many more duty cycles for the system to deal with.

Ultimately, the constraint of a system can be described by a triangle whose boundaries are defined by the number of transmit lines that can be fired. Lines widely spaced can increase the volume size at the cost of lowering the resolution for a constant frame rate (Figure 3-11). Tight line spacing can be used in zoom modes to increase resolution, but at the price of a smaller volume. The number of transmit lines is a key determinant of frame rate; more lines increase the resolution but lower the frame rate.

Figure 3-11 Relationship among frame rate, resolution, and sector size in three-dimensional beamforming. For a fixed number of lines, the sector size can be increased but at a tradeoff of resolution, which makes the point spread function more “blobby.” Firing more scanning lines can increase resolution, sector size, or both (depending on line density), but firing more lines reduces frame rate.

Beamforming can be subdivided into a coarse stage that occurs in the system and a fine steering or micro-stage that occurs in the transducer. The act of combining element signals is known as summing. Summing of per-element pulses is what ultimately creates a scan line. The general sequence of events is as follows:

The most significant aspect of a 3D beamformer is the ability to steer both azimuthally (laterally) and elevationally. This creates a spherical coordinate system. The limits of resolution stem from lateral or elevational line spacing. Samples within a line are more finely spaced in the radial dimension (i.e., within a scan line) and are farther apart across scan lines in the lateral or elevational directions. Therefore, line spacing is a fundamental determinant of image quality.

Gating refers to the act of acquiring multiple acquisitions (timed to the ECG) to combine them at some later point. 3D acquisitions in the 1990s entailed acquiring 60 to 180 2D slices and interpolating them to create a 3D reconstructed volume. Since this process took several minutes, it was prone to misalignment and inferior resolution. Gating is still used today to overcome limitations in the speed of sound by combining only a few 3D slabs (e.g., 4 to 8) to create a larger volume. This process takes only seconds. Moreover, color Doppler techniques depend on analyzing multiple transmit events fired along a single line. The multiple returning pulses are compared or cross-correlated to infer velocity. Nevertheless, gating is prone to error in patients with irregular rhythms such as atrial fibrillation or multiple ventricular premature contractions.

3D ultrasound imaging is still subject to the laws of acoustic physics even though it is ungated or “live 3D.” First, the aperture is an important determinant of image quality. Larger apertures allow better beamforming from a focus point of view. Unlike abdominal imaging, transthoracic echocardiography is limited by the ultrasound aperture that can fit between a rib space. In TEE, the aperture is limited by dimensions that can be accommodated within an adult esophagus. TEE generally uses higher frequencies compared with its transthoracic counterparts. This allows better image resolution for a given aperture size. (The higher the frequency, the better the ultrasound beam can focus, but this comes at the loss of penetration.) Since TEE is not limited by significant chest wall acoustic aberration, the higher frequencies and better acoustic substrate allow higher resolution compared with transthoracic imaging (Figures 3-11 and 3-12).

Figure 3-12 Three-dimensional (3D) volume rendering of mitral insufficiency. The left ventricle is shown, and color Doppler data have been scanned in 3D coordinates to the mitral valve leaking. Aliasing is also shown between red and blue pixels as the jet velocity exceeds the Nyquist sampling limit to quantify its true velocity.

Artifacts such as shadowing, reverberations, multipath transmission and aberration play a role in degrading the ultrasound image. As with 2DE, 3D ultrasound cannot image through metallic or highly calcified objects. One of the most significant issues pertains to the quality of the image as seen through chest wall windows versus the esophagus. This is due mainly to two phenomena: aberration and multipath reflection. Aberration stems from wavefronts traveling through different media with different velocities of sound. Fascial tissue layers create a distortion of the traveling wavefronts. This can be corrected for, in a limited way, by accounting for the varying speeds of sound. The layers of the chest wall contain varying degrees of adipose and connective tissues. This creates not only aberration but also multipath degradation. As ultrasound waves get diverted to altering paths of propagation, the superposition of transmitted and returning echo signals consists of wavefronts of both desired and nondesired targets. Unwanted, but real, signals are termed clutter. Since the esophagus represents a thin wall consisting of a stratified squamous epithelium and smooth muscle, these ultrasound effects do not occur in 3D-TEE. Hence, image quality is higher, and clutter is lower.

Quantification for Three-Dimensional Echocardiography

Once 3D data are processed, they are “volume rendered” to create the appearance of three dimensions (Figures 3-13 to 3-16). 3D color voxels can be rendered as well (see Figure 3-12). The considerable advantage of 3DE in acquiring ultrasound images of the beating heart makes it especially useful for cardiac quantification. While the display of anatomy in its true 3D state is important, many physicians believe that the single most significant value that 3DE has is its ability to quantify. Myocardial and valvular motion occurs in three spatial dimensions. Traditional 2D scanning planes do not capture this entire motion, or the plane can “slip” during scanning. Quantifying requires segmenting or separating out structures of interest from the 3D image. Machine vision techniques use methods that define an interface, such as a left ventricular (LV)-endocardial border. This interface is generated as a mesh of points and lines and displayed by a process known as surface rendering (Figure 3-17).

Figure 3-13 Transthoracic imaging showing apical four-, two-, and three-chamber views of the left ventricle. This is an example of volume rendering in which three-dimensional (3D) volume elements known as voxels are used to create the 3D appearance scanned by a 3D acoustic sweep. In addition, hue (color) is used to add to the visual effect of depth. Ultimately, these voxels are converted to two-dimensional (2D) picture elements (pixels) in a process known as volume rendering. Even though the picture is in 2D, real 3D data are used to create these images.

Figure 3-14 Three-dimensional transesophageal echocardiography of the mitral valve showing a flail segment near P2.

Figure 3-15 Three-dimensional transesophageal echocardiography image of the mitral valve showing the effect of cropping data. The upper left panel shows an uncropped image looking down from the left atrium. 1, A cropping plane has been used to cut away data along the green axis. This allows better visualization of the anterior and posterior mitral leaflets. 2, Another cropping pane, in this case along the red axis, has been used to cut away anterior anatomy as part of the aortic valve. In this case, since the cropping was at the side, there is not much benefit. 3, A cropping plane along the blue axis has been used to crop away data from the left atrium toward the top of the image.

Figure 3-16 Multiplanar reformatting of two-dimensional (2D) slices taken from a three-dimensional (3D) dataset. Three orthogonal 2D views can be extracted from the 3D dataset (lower left). This allows any plane orientation of 2D views in displaying anatomy of interest.

Figure 3-17 Surface rendering and segmentation of mitral leaflets. A prolapsed segment of P3 is shown in red. This portion of the leaflet extends above the annular surface into the left atrium. Also shown is the regurgitant orifice caused by malcoaptation of the leaflets. A, anterior; AL, anterolateral; Ao, aortic valve annulus; P, posterior; PM, posteromedial.

The application of 3D beamforming in a TEE probe allows visualization of the beating mitral valve. The mitral valve can be segmented with significant accuracy by using this approach. The true 3D structure of the mitral annulus, leaflets, and chordal apparatus can be measured. This further allows sophisticated analyses of the nonplanar shape of the mitral annulus.^13–24 These 3D measurements include the following:

The ability to acquire structural data of the mitral valve allows advanced engineering techniques to examine the nature of fluid flow in three dimensions. For example, the proximal isovelocity surface area (PISA) technique is an important quantitative tool to estimate the degree of mitral regurgitation. It depends, however, on geometric assumptions, which must be respected.²⁵ By using actual echocardiographic data from a mitral valve that has been segmented after 3DTEE, the flow of blood through the orifice can be simulated to see the nature of converging isovelocity fluid zones. Near an irregularly shaped orifice, the zone is necessarily irregular and nonspherical. This indicates that care should be exercised in selecting the PISA aliasing velocity to select a near-hemispherical convergence zone (Figure 3-18).²⁶

Figure 3-18 Three-dimensional (3D) computational fluid dynamics flow simulation of proximal isovelocity surface area using 3D transesophageal echocardiography segmentation of a regurgitant mitral orifice. Upper left, using similar technology as in Figure 3-17, a 3D mesh is created from 3D data and imported into a finite element environment (upper right). After setting initial pressure conditions, 3D stream lines pass through the regurgitant orifice from top to bottom (bottom left) (valve is upside down as from the left ventricular point of view). Close to the orifice, finite element simulation indicates that the proximal isovelocity convergence zone is not necessarily hemispherical (bottom right).

3D has a significant benefit over 2D in that if the entire LV chamber is encompassed in the acquisition volume, no foreshortening errors or assumption of LV volume are created. Biplane and triplane methods help avoid foreshortening errors and benefit from higher frame rates than 3D live acquisitions; however, if an aneurysmal dilation occurs between planes, the computed LV volumes will have some interpolation error. Surface rendering of the LV is used for chamber quantification. This can be used on all four chambers. Computer techniques allow calculation of volume, regional wall motion, and regional synchrony.^27–43 Specifically, a 3D deformable model is used to find the LV endocardial surface in three dimensions. This is the most accurate way to quantify LV volumes. Moreover, 3D LV remodeling can be parametrically displayed using differential geometry techniques.⁴⁴ This is done by calculating local curvatures around the LV endocardium (Figure 3-19). Since the normal LV is not a sphere, local radii of curvature vary around the segments of the LV. For example, the normal septum has different geometric characteristics from those of the apex. Quantification of LV synchrony is possible in 3D as well. The required frame rate depends on the questions being asked. Frame rates of 30 Hz (33 ms between frames) are inadequate to quantify intramyocardial motion; these are better suited to be studied by tissue Doppler or speckle tracking techniques. However, regional synchrony can be measured by 3D endocardial excursion because it assesses blood ejection, not tissue motion. 3DE provides a more complete assessment of 3D wall motion. Challenges in frame rate limitations, however, must be taken into account. Assessment of global function is more tolerant of lower frame rate acquisitions compared with that of regional function. That is, to find the ejection fraction, only accurate estimates of end-diastolic and end-systolic volumes are needed (Figure 3-20).

Figure 3-19 The left ventricular (LV) segmentation mesh technique (shown in Figure 3-19) can be used to create parametric displays of regional LV remodeling. At left is a normal patient with a smaller endocardial cavity and inward curvature of the LV septum. Patients with dilated cardiomyopathy (DCM) lose septal curvature and also have concomitant apical dilation (right). The histogram of curvatures has a smaller standard deviation and hence is more spherical in character. (A sphere has zero standard deviation since it has uniform radius.) Neg, negative.

Figure 3-20 Three-dimensional segmentation of the left ventricle. The bull’s-eye shows regional segmentation by anatomy. Both global and regional function without geometric assumptions can be made with this approach. Regional ejection curves for each segment can be computed over the cardiac cycle. All the regional waveforms appear similar (are correlated) and indicate synchronous ejection in this patient.

Acknowledgment

The author gratefully acknowledges the invaluable transducer engineering input from Michael Peszynski.