Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
3D Echocardiographic Image Acquisition, Display, and Analysis
Three-dimensional (3D) echocardiography (3DE) is one of the most important innovations in the field of cardiovascular ultrasonography. To maximize the information obtained from a 3DE data set, image acquisition and analysis should be performed with a deep understanding of the underlying technical principles and using a systematic approach. Although there are differences among cardiovascular ultrasound equipment manufacturers with respect to terminology and technical performance, the underlying fundamental principles among vendors remain the same. The goal of this chapter is to provide a practical guide that focuses on the core principles required to acquire, display, and analyze cardiac structures using 3DE and describes the limitations of this imaging modality. The first part of this chapter describes the physics, technical factors, and terminology related to 3DE. The second part provides a practical description of the clinical use of 3DE with respect to acquisition, display, and analysis of specific cardiac structures.
Ultrasound Imaging System Function
The basis for ultrasound image formation is as follows. To acquire an image, the transducer transmits a high-frequency sound beam. When this beam hits an acoustic interface, such as an intertissue boundary, some of the sound waves are reflected and “echoed” back to the transducer. An image is then created by incorporating information regarding the distances and intensities of the echoes transmitted to the probe from the tissue boundaries based on calculations that incorporate the speed of sound in the tissue and the time required for each echo to return. In the transducer, the piezoelectric element is the component that transmits and receives ultrasound information and transforms it into electric signals, which are then sent to the ultrasound machine. A single-element transducer can acquire and form images only in a single spatial and one temporal dimension. In practice, this is M-mode imaging in that only structures in the beam line are visualized.
Two-dimensional (2D or B-mode) images are acquired using linear phased array transducers in which 48 to 128 piezo elements are arranged in a single row (one-dimensional linear array) with each element functioning separately ( Fig. 1.1 ). (2D images can also be acquired with mechanical transducers, which consist of a rapidly moving single crystal.) Each element is activated according to a specific sequence with a delay in phase with respect to the transmit initiation time ( Fig. 1.2 ). The individual waves generated from each element interact constructively and destructively to form an overall wave that has a direction and is known as a radially propagating scan line. Because the elements are arranged in a single row in a 2D linear phased array transducer, the transducer can steer and focus in only two directions ( Fig. 1.3 ), axial and azimuthal (lateral). When the 2D image is formed, the resolution in the elevation plane is fixed by the vertical dimension of the elements, which restricts the slice thickness.
In an ultrasound machine, the transmitting and receiving functions create scan lines through beamforming and summing ( Fig. 1.4 ). Beamforming or spatial filtering is a signal processing technique that uses the array of elements to create directional or spatial selectivity of signal transmission and reception. The time delay before the activation of each element and the strength of the sent signal allow focusing and steering of the transmission beam. Summing is the act of combining signals by summing the pulses from each transducer element to create a scan line. For the ultrasound machine to perform beamforming, the following components are required: high-voltage transmitters, low-noise receivers, analog-to-digital converters, digital delay lines, and delay controllers. In 2D imaging, beamforming does not occur in the transducer but in the ultrasound machine, consuming approximately 100 W of energy and 1500 cm 2 of a personal computer electronic board area.
With 3DE, the transducer elements are organized in a matrix formation (2D array). Early 3D transducers were called sparse arrays because not all the elements in the matrix were simultaneously electrically active. These were the first transducers to enable acquisition of 3D images that could be immediately visualized, but because of the sparse array activation pattern, control of the ultrasound beam was not precise, and diffraction effects such as grating lobes were commonly encountered. The 3D transducers used today are fully sampled matrix array transducers in which each element is simultaneously electrically active. , 3D matrix array transducers are composed of 2000 to 3000 piezoelectric elements with operating frequencies ranging from 1 to 5 MHz for transthoracic and 2 to 8 MHz for transesophageal transducers. , Phasic activation of the elements in the matrix array allows 3D transducers to generate a scan line that propagates radially (axial direction) and can be steered in two directions (azimuth/lateral and elevational), creating a pyramidal coordinate system (see Fig. 1.3 ).
As mentioned previously, in a 2D probe, all components required for beamforming are contained in the ultrasound machine. To maintain this configuration with a fully sampled 3D matrix array probe containing 3000 elements would require a 3000-channel system and cable, 4 kW of power consumption, and a large personal computer electronics board area to contain all the circuitry, which would not be practical. A major hurdle in the development of fully sampled matrix array probes was the need to reduce power consumption and the size of the connecting cable while maintaining the electric interconnections for every element, to ensure that each element remains independent with respect to transmitting and receiving. This was achieved through placement of specifically engineered miniaturized application-specific integrated circuit (ASIC) boards within the transducer. The first commercial, fully sampled matrix array transducer had 24 to 26 ASICs, which were connected to the approximately 3000 elements. This allowed the 3000 elements to be independently active while simultaneously keeping the size of the transducer cable reasonably small. With further miniaturization advancements, current transthoracic and transesophageal probes have a single ASIC.
As a result of the placement of the transducer ASICs, beamforming was split into two components (see Fig. 1.4 ): a microbeamforming stage that uses the beamforming circuitry in the transducer and requires less than 1 W of power and a coarse stage of beamforming that occurs through conventional cables in the ultrasound machine. , When microbeamforming occurs in the transducer, it is no longer required that each piezoelectric element be connected to the ultrasound machine. The 3000-channel circuit boards within the transducer control the fine steering by delaying (through fine circuitry integrated into the ASIC) and summing signals within subsections of the matrix, known as patches. This allows the number of cables connecting the transducer to the ultrasound machine to be reduced from 3000 to between 128 and 256. Coarse steering occurs in the ultrasound machine, where digital delay lines convert the analog signals to digital signals.
However, placing the circuitry for microbeamforming in the transducer head generated significant heat proportional to the mechanical index used during imaging. Two different strategies have been taken to address this issue: (1) active cooling, in which the heat is actively transported through the transducer cable, and (2) passive cooling, in which heat production is reduced through improved crystal manufacturing processes that create single-crystal materials with homogenous solid-state domains and piezoelectric properties. These new crystal materials improved the conversion of transmitted power into ultrasound energy and that of received ultrasound into electrical energy. This increase in efficiency during transduction resulted not only in reduced heat production but also in wider bandwidths, increased echo penetration and resolution, improved image quality, reduced artifacts, reduced power consumption, and increased Doppler sensitivity.
More recent developments have resulted in improved side-lobe suppression, increased sensitivity and penetration, and the implementation of harmonic capabilities that can be used for both grayscale and contrast imaging. This has led to significantly smaller matrix transducers with improved image quality and has created the capability for a single transducer to acquire both 2D and 3D images.
Resolution in 3D Echocardiography
In 3DE, spatial and temporal resolution are inversely linked. Optimizing image quality for either spatial or temporal resolution will adversely affect the quality of the other.
Spatial Resolution
The spatial resolution of an imaging system is characterized by its point-spread function. The point-spread function is the image obtained of an infinitesimal point object. A single pixel in an idealized image (point input) will be reproduced as something other than a single pixel in the obtained image. This degree of blurring (spreading) of any point object varies according to the dimensions employed.
Each point in a 3DE pyramidal data set is located using three coordinates (see Fig. 1.3 ): (1) depth within the 3D volume, which is the distance from the transducer (x-axis); (2) the azimuth plane (y-axis); and (3) the elevation plane (z-axis). Spatial resolution is different along each of these axes and is related to the spacing between scan lines. 3D matrix array transducers generate scan lines that propagate radially (axial direction) and can be steered in two directions: azimuth (lateral) and elevational. Samples within a scan line are more finely spaced, whereas scan lines in the other directions (i.e., elevational or lateral) are farther apart. Overall, the elevational plane has poorer image resolution because the scan lines are farthest apart in this plane.
Generally, spatial resolution along the x-axis is similar to that of 2D scanning, which is one-half wavelength. In the other two planes, it is twofold to threefold worse. The spread in current 3D imaging systems is 0.5 mm in the axial (x) dimension, 2.5 mm in the lateral (y) dimension, and 3 mm in the elevation (z) dimension. Therefore, the best images with minimal blurring are obtained in the axial dimension and the worst ones in the elevation dimension. The portion of an object or region that is perpendicular to the ultrasound beam will be best displayed in the axial resolution, and it will have higher resolution than the region that is parallel (see Fig. 1.3 ).
In practice, for transthoracic imaging this means that the poorest 3D images are obtained from the apical approach, in which cardiac structures are predominantly imaged using the lateral and elevation dimensions. Conversely, the best 3D images are obtained from the parasternal approach, which uses the axial and lateral dimensions. For this reason, when using transesophageal echocardiography (TEE), the mitral valve (MV) is visualized with better resolution than the aortic valve (AV).
Temporal Resolution
A major limitation to 3D imaging, and echocardiography in general, is the constant speed of sound in myocardial tissue and blood (1540 m/s). Because of this limitation, the maximum number of pulses that can be transmitted per second without creating interference is obtained by dividing the speed of sound in myocardial tissue and blood by the imaging depth or distance that the pulse has to travel back and forth. The maximum number of pulses limits the number of 3D pyramidal volumes per second that can be imaged given a desired pyramidal size and spatial resolution. Therefore, in 3D imaging, an inverse relationship exists between volume rate (temporal resolution), volume size, and spatial resolution (number of scan lines). However, manufacturers have developed solutions to circumvent the limitation of fixed sound speed. These methods include electrocardiogram (ECG)–gated stitching of subvolumes (multibeat), a real-time zoom feature that reduces the field of view, parallel receive beamforming that allows true real-time large pyramidal (full-volume) acquisition, interpolation of images or image sectors, frame (volume) reordering, virtual array, multiline transmission, and high-pulse repetition frequency. Of these methods, ECG-gating and real-time zoom are options that can be adjusted by the operator. The other methods are either built into the ultrasound system or require post-acquisition software. Each of these methods is discussed in this section except for real-time zoom, which is a pyramidal size option that is discussed in the section on 3D acquisition modes.
Single-Beat or Multibeat Acquisition
Single-beat 3D imaging refers to the acquisition of multiple pyramidal data sets per second during a single heartbeat ( Fig. 1.5 ). This methodology overcomes the limitations imposed by rhythm disturbances or respiratory motion. Historically, its use was limited by poor temporal and spatial resolution; however, technological advances allow current ultrasound machines to acquire 3D data sets of large structures with high temporal resolution using single-beat imaging ( Table 1.1 ).
TABLE 1.1
Strengths and Weaknesses of Single-Beat Versus Multibeat 3D Image Acquisition.
| Single-Beat | Multibeat | |
|---|---|---|
| Artifacts | None | Stitch artifacts from motion or cardiac arrhythmias |
| Temporal resolution | Lower | Higher |
Despite temporal and spatial resolution improvements with single-beat acquisition, multibeat 3D acquisition is still important in patients with structures that require large pyramidal volumes and high temporal resolution. Multibeat acquisitions are pyramidal acquisitions of multiple narrow volumes of data over several heartbeats (ranging from two to six cardiac cycles) that are subsequently stitched together to create a single volumetric data set. In this way, pyramidal volume size can be maintained while the volume rate is increased by obtaining and stitching these subvolumes together. However, multibeat imaging is inherently susceptible to stitch artifacts, which appear as subvolume misalignment and are created by patient or transducer motion, cardiac translation during respiratory or cardiac motion, or changes in cardiac cycle length.
Parallel Beamforming
Parallel beamforming is a method wherein the system transmits one wide beam and receives multiple narrow beams in parallel; this increases the volume rate by a factor equal to the number of receive beams ( Fig. 1.6 ). This parallel processing of the received data allows multiple scan lines to be sampled in the same amount of time that a conventional scanner would take to create a single line. It is achieved by having each beamformer focus along a slightly different direction that was insonified by the broad transmit pulse. However, as the receive beams are steered farther and farther away from the center of the transmit beam, lower energy signals are received, resulting in a reduction of signal strength and resolution. Increasing temporal resolution by increasing the number of parallel beams requires an increase in the size, cost, and power consumption of the beamforming electronics, thereby decreasing the signal-to-noise ratio and the contrast resolution.
Interpolation of Images or Image Sectors
Interpolation is a function in which unknown points are estimated by using known data. In 3DE, this occurs by interpolating results for either image subsections or frames between those acquired. When interpolation occurs within a matrix array, the array functions like a checkerboard ( Fig. 1.7 ). During the first cardiac cycle, some elements are active, and the nonactive elements are interpolated. During the next cardiac cycle, the initially nonactive elements become active, and their results are compared with the interpolated image. If the data acquired by the initially nonactive elements improves the 3D image, those data are used. Otherwise the interpolated data remains in use. Limiting the number of active elements results in faster processing, allowing more volumes to be acquired over a given period of time, thereby increasing the temporal resolution. Similarly, interpolation can be used to construct intermediate volumes from adjacent volumes.
Frame Reordering
In frame reordering, a high frame (volume) rate algorithm is used. It reorders 3D volumes of a periodically moving cardiac structure taken at a number of instants over several cardiac cycles. All of the volumes used are acquired ad hoc, without ECG-gating at different time periods. This results in faster acquisition times (tenths of a second) and volume rates that are higher than those typically achieved by the ultrasound system during a single cardiac cycle. The algorithm functions by identifying the temporal sequence of volumes on either side of the ECG R wave and interleaving these volumes into a coherent ordering with respect to the R-wave peak. By reordering only those frames close to the QRS, the beat-to-beat time interval variability is removed. This results in an increase in 3D volume rate from 10 to 540 frames per second (fps). Weaknesses inherent to this method include the 20-minute processing time per sequence, low beat-to-beat variability of the heart or cardiac structure of interest within the QRS complex, and other factors that affect temporal coherence. An advantage of this methodology is that it can be performed with a standard 3DE ultrasound system and a standard computer.
Additionally, the same algorithm can be used to reorder subcomponents of the 3D volume to improve image quality. In 3DE, different parts of the volume are acquired at different times because of the manner in which scan lines are sent, read, and composed into a Cartesian volume format. Because the frames are not simultaneously acquired, it is possible for individual voxels in a frame to be slightly out of sequence. Reordering these voxels or a subcomponent instead of whole frames can be a solution to this problem.
Virtual Array
Current commercial 3DE ultrasound systems typically rely on hardware-based focused ultrasound beams, limiting the volume rate to a few tens of volumes per second. In contrast, high-volume-rate (ultrafast) 3DE images are those in which thousands of volumes are acquired per second. This is achieved through the transmission of a small number of defocused ultrasound waves that insonify the entire volume of interest. With this method, dynamic focusing is performed in the receive mode. During transmit mode, the coherent synthetic summation of the ultrasonic volume acquired for each transmission occurs, allowing for the restoration of a dynamic focus without compromising the ultrafast volume rate. This is achieved through the use of a virtual array located behind the probe. ,
The virtual array is used to synthetically form the entire imaging volume ( Fig. 1.8 ). For each individual virtual source, delays are computed and a sub-aperture is defined. When the virtual sources are located far behind the probe, a larger sub-aperture is present; this results in greater emitted energy and a smaller field of view. When virtual sources are located near the physical probe, the sub-aperture used is smaller, and the curvature of the emitted waveform is increased. This results in the insonification of a large field of view at the cost of a lower propagated energy. In the extreme case of sources located at infinity behind the probe, tilted plane waves are obtained.
Although this method allows for a high volume rate, limitations still exist with respect to the balance among contrast, resolution, volume rate, and field of view. Additionally, the use of defocused waves may result in poor image quality because of the wide lateral spreading of the transmitted energy, possible high side lobes and motion artifacts, and inability to generate second harmonics. However, this method holds promise because it allows mapping of blood flow and tissue motion in the entire 3D field of view.
Multiline Transmission
Multiline transmission is a technique in which multiple ultrasound pulses focused along different steering directions are transmitted simultaneously. The gain in volume (frame) rate is equal to the number of beams. Traditional multiline transmission is susceptible to crosstalk artifacts, which result from the interactions among the ultrasound fields on transmit and receive waveforms. This can be eliminated by simultaneously transmitting beams along the transverse diagonal of the transducer.
High Pulse Repetition Frequency
This method increases color Doppler volume rates by combining multiple narrow beams with high pulse repetition frequency (PRF). , With high PRF, pulses are transmitted with three times the frequency that is needed to allow the echo from the farthest depth to return. For example, if two pulses are emitted, the echo from the first pulse will return from the farthest depth at the same time the echo from the second pulse returns from an intermediate depth. When this occurs, there is no way to determine whether the signal originates from the deepest level or the intermediate level. The high PRF increases the Nyquist limit, thereby shortening the acquisition time. This method has been demonstrated to have high accuracy compared with cardiac magnetic resonance in quantifying aortic regurgitation.
Relationship Between Spatial and Temporal Resolution
Overall, the relationships among volume rate, number of parallel receive beams, sector width, depth, and line density are described by the following equation:
Volume rate = (1540 × Number of parallel receive beams)/[2(Volume width/Lateral resolution) 2 × Volume depth]
Thus, on an ultrasound machine, volume rates can be increased by decreasing volume depth or width or by reducing the number of parallel receive beams ( Fig. 1.9 ). However, adjusting the parallel receive beams affects the signal-to-noise ratio, altering image quality. Similarly, changing the scan-line density in the pyramid can alter lateral resolution and affect image contrast.
Multiview 3D Echocardiography Image Fusion
Whereas the discussions here involve 3DE images that are obtained using pyramidal volumes, 3D images can also be generated by fusing multiple 2D images. Typically, these image fusion systems require post-acquisition reconstruction of 2D images acquired with the use of a 2D transducer that is tracked in three dimensions by a spatial tracking system. , This was the methodology used initially to create 3DE images, but it is now used to improve image quality of structures that are challenging to image, such as the right ventricle.
3D Echocardiography Data Acquisition
Overall, 3DE data acquisition can be divided into four main steps ( Fig. 1.10 ). First, the 2D image should be optimized. Second, based on the structure evaluated, a choice should be made with respect to acquisition viewing mode (multiplane or pyramidal volume) and volume size. These choices will be affected by whether 3D color Doppler images will be obtained. Subsequently, decisions should be made to optimize volume rate (e.g., single vs. multibeat, scan-line density adjustment). Third, depending on the structure of interest, the 3D data set may be cropped and the image optimized. Fourth, the optimal method for displaying the 3D data is determined, and, if needed, the 3D data should be analyzed.
2D Image Optimization
2D image quality should be optimized before acquisition of the 3DE data set. Acoustic shadowing and other noise on the 2D image will also appear within the 3D data set. Image optimization can be achieved by using breathing maneuvers to minimize transducer motion and through adjustments to regional and global gain. Most importantly, when setting up the 3D acquisition, multiplane visualization should be used to ensure that the planes perpendicular to the original reference imaging plane are also optimized. This ensures that the entire structure of interest will be included in the 3D pyramid.
3D Acquisition Modes
Simultaneous Multiplane Mode
Simultaneous multiplane imaging permits visualization of two to three planes in B-mode or color flow Doppler echocardiography. The first image is typically a reference view of a particular structure, whereas the other images represent planes rotated at any angle from the reference plane ( Fig. 1.11 and Table 1.2 ). Multiplane imaging in the elevation plane is also available.
TABLE 1.2
Uses of Multiplane Mode.
| Pre-acquisition |
|
| Post-acquisition |
|
| Analysis |
|
| Interventional procedures |
|
Pyramidal Size
As described previously, pyramidal size affects spatial and temporal resolution. Ideally, the smallest pyramid that captures the information required should be used. Although original 3DE acquisition was limited to three relatively fixed pyramidal sizes, these and other options are now available on ultrasound equipment as starting points for further adjustment of the pyramidal size. All of these sizes can be acquired using single-beat or multibeat approaches to increase spatial and temporal resolution. Three basic pyramidal sizes are briefly described here ( Fig. 1.12 and Table 1.3 ).
TABLE 1.3
Choice of 3D Pyramidal Volume Size Based on Cardiac Structure.
| Pyramidal Size | Cardiac Structure | Strengths | Weaknesses |
|---|---|---|---|
| Zoom |
|
|
|
| Narrow sector |
|
|
|
| Wide sector |
|
|
|
Zoom
Zoom is typically the smallest pyramidal size available for acquisition (∼30 by 30 degrees) and is used for small cardiac structures. It increases temporal and spatial resolution by decreasing the region of interest in both the azimuth and elevation planes. However, there can be a loss of spatial resolution if the sector is magnified greatly.
Narrow Sector
Depending on the vendor, the narrow sector is also called bird’s eye or live mode . Historically, this size permitted real-time display of a pyramidal volume of 30-to-50 degrees by 60-to-90 degrees. This mode allows imaging of a narrow region of the structure of interest plus the surrounding areas, so is ideal for use during percutaneous procedures.
Wide Sector
Depending on the vendor, the wide sector is also called full volume or large mode. This pyramidal size is the largest acquisition sector possible and is best used for structures such as the left or right ventricle. The wide sector data set can be cropped or multiplane transected to remove tissue planes; this allows one to identify or extract components of valvular structures within the volume or to visualize 2D cross-sectional x, y, and orthogonal planes using off-line analysis software.
Color-Flow Doppler
Currently, 3DE color Doppler imaging can be acquired with one to six individual gated volumes. Volume rate differences between color Doppler and noncolor 3D volumes that are otherwise acquired with identical parameters may or may not be significant depending on the ultrasound machine vendor. Traditionally, there is a reduction in volume rate when 3D color Doppler imaging is used.
Challenges with 3D Echocardiography Acquisition
Temporal Versus Spatial Resolution
The main trade-off in 3DE imaging is between temporal resolution (volume rate) and spatial resolution ( Table 1.4 ; see also Fig. 1.9 ). As discussed previously, improvements in spatial resolution can be achieved by increasing the scan-line density of the 3D volume. However, with this method it takes longer to acquire and process the image, limiting the overall volume rate. By decreasing the 3D volume size, volume rates can be increased while maintaining spatial resolution. Volume rates can also be increased through the use of ECG-gating (multibeat acquisition). To prevent artifacts secondary to ECG-gating, ultrasound companies have developed technology and methods that allow acquisition of single-beat, wide-sector 3D pyramidal data with high spatial and temporal resolution.
TABLE 1.4
Choice of Temporal and Spatial Resolution Based on the Cardiac Issue Studied.
| Issue | Requirements | Reason | Possible Solutions |
|---|---|---|---|
| Left and right cardiac chamber volumes | High temporal resolution with a large pyramidal size | To capture true end-diastole and end-systole | Multibeat acquisition High-volume-rate methods (e.g., parallel beamforming, virtual apex) Decreased line density |
| Small, highly mobile cardiac structure (e.g., valves, vegetation) | High spatial and temporal resolution | To visualize small mobile structures | Decreased pyramidal size Use high-volume-rate method (e.g., multibeat) Increased line density |
ECG-Gating and Breath-Hold
ECG-gated data sets are challenging in patients who are unable to hold their breaths or have arrhythmias, as this can result in stitch artifacts (see Fig. 1.5 ). Gating artifacts are best identified using multiplane viewing. Typically, subsectors are acquired in sweep planes parallel to the reference image. Thus, stitch lines are parallel to the reference imaging plane and therefore can only be identified from planes perpendicular to the sweep/reference plane. Gating artifacts can be also minimized by optimizing the ECG tracing so that a distinct R wave is present and by minimizing patient and probe motion.
3D Image Optimization
The goal of 3D image optimization should be to adjust the 3D rendering of the data set to give it a 3D appearance while simultaneously removing artifacts. Typically, these adjustments can be performed before or after the 3D data set acquisition. The settings available vary depending on the ultrasound vendor. However, core settings available across vendors that may need adjustments include gain, brightness, smoothing, colorization, and compression ( Table 1.5 ). In practice, with current ultrasound systems, gain is the setting that most often requires adjustment because available presets provide adequate image quality.
TABLE 1.5
Challenges with 3D Image Optimization Options.
| Option | Increase | Decrease |
|---|---|---|
| Gain | Loss of resolution Loss of 3D perspective/depth |
Dropout |
| Brightness | White out, loss of depth | Image too dark |
| Smoothing | Loss of texture/details of the structure under study | Overidentification of abnormalities |
You're Reading a Preview
Become a Clinical Tree membership for Full access and enjoy Unlimited articles
If you are a member. Log in here