Chapter 43: VOLUME SONOGRAPHY: CORE CONCEPTS FOR CLINICAL PRACTICE

Conventional high-resolution two-dimensional ultrasound (2DUS) is commonly used for numerous indications in obstetrics and gynecology and is widely considered a clinically useful technique. This imaging technology requires the operator to scan through the patient's anatomy with a 2D ultrasound beam to obtain cross-sectional planes. The 2D imaging planes are typically thin and only display cross–sectional information. The ultrasound operators typically put in a lot of effort and time to mentally develop a three-dimensional (3D) concept of the region of interest with depth and perspective. In addition, several image planes are either very difficult or inherently simply impossible to be obtained with the conventional 2DUS techniques.

Some of the early research works on 3D volume acquisition and display methods started in 1970.¹ Early prototypes had sophisticated setups for that time, with static arms, which were later equipped with position sensors to more reliably register the acquired 2D data in time and space. Only in the early to mid-1990s did the technology really start to leave the industry's research and development labs and become a more appealing clinical tool.^{2,3,4,5, and 6} Significant developments in transducer technologies, and advances in computer signal processing and image display finally made possible reliable techniques to acquire and visualize ultrasound volumes.⁷

Three-dimensional ultrasonography (3DUS) can depict the anatomy of the region of interest as static or even in motion. The displays have become fully interactive environments that allow the operator to have full control over the scan planes and evaluate the anatomy in user-selectable uniplanar, biplanar, orthogonal, and multiplanar displays, as well as rendered 3D views. Volume imaging techniques are not new. Radiologists have been acquiring and displaying computed tomography (CT) and magnetic resonance (MR) slices in a volume for years. What is unique to 3DUS volume acquisition and display technologies is that they enable classic volume imaging methods to be used in a much smaller anatomical scale, even intracavitary, with the known efficiency and economics of ultrasound.

In the literature, the terms 3D ultrasound and volume ultrasonography are commonly used interchangeably. They both will also be used here, understanding that they represent the same concept. In 3DUS, a volume (rather than a slice) of ultrasound data is acquired and stored. The stored data can be visualized, reformatted, and analyzed in a variety of ways. The volume data can be quantified. For the first time we can reliably measure volumes of any organ system, provided they can fit in the volume acquisition box and the organ's borders have good interfaces. This information can be digitally saved and retrieved at any later time. Small movies of these volumes at different projections that enhance depth perception can be created and stored as well.

The clinical applications of 3DUS in obstetrics and gynecology have been explored in numerous published works.^{8,9,10,11,12,13,14,15,16,17,18, and 19} Although there has been much more extensive research into the obstetrical applications of 3DUS, there is increasing evidence in the literature that 3DUS does have very substantial applications in gynecology as well.^{10,20,21,22,23, and 24.}

Volume data in 3DUS represent the acquisition of a number of 2DUS data and an extension of conventional technique. As such, 3DUS does not replace 2DUS, but rather extends its capabilities. Using 3DUS for clinical applications initially may require an additional investment of time (typically 3 to 5 minutes). With some applications, this time depends on the complexity of the case and experience of the examiner. The additional time is not as much related to scanning and volume acquisition as to exploring the saved volume data. As a result, although the total examination time may be longer due to the amount of information to be processed, the scanning time, or the time the patient is exposed to the ultrasound beam, with experience should be reduced. During the examination itself, the acquired volume is only briefly scanned to ensure that the region of interest has been adequately included and there are no artifacts present. If so, the volume is saved in the hard drive of the machine or is sent to an archive server or picture archive and communication system (PACS). Exploration of the saved volumes can be performed after the patient leaves the room or is discharged. A typical sonographic examination workflow using a 3DUS system involves (1) scanning the anatomy using conventional real-time 2DUS, mapping the position and orientation of the organs to be visualized relative to the axis of the beam, and obtaining standard image planes for volume acquisition; (2) acquiring volumes either slow or fast (depending on whether it is a moving target or not) in specific acquisition planes known to yield the required planes; (3) volume exploration for the artifact-free presence of the expected image planes; (4) quantitative and qualitative analysis of stored volume data such as multiplanar analysis, rendering, calculating volumes, and creating cine loop clips if need be; and (5) storage of volume data in the hard drive (Figure 43-1).

A common question among ultrasound operators is why 3DUS should be used in a particular clinical scenario. Experienced 2DUS operators especially have to convince themselves why they would need a 3DUS scan in addition to 2DUS, which has been successfully used for many years. If 3DUS proves to be useful, these individuals will have to put a significant amount of effort into learning about, purchasing, and using 3D equipment in everyday clinical practice.

There are 2 main reasons why a clinician would need to consider the use of 3DUS (Figure 43-2):

1. Volume ultrasound can obtain image planes that are either impossible or very difficult to obtain.

2. Volume ultrasound can provide depth, which is impossible with 2D ultrasound.

In everyday clinical work, an ultrasound operator is required to obtain the so-called standard ultrasound views and extract clinically useful information based on them. These standard views have been studied in the past for their ability to be acquired by most operators with adequate training. They also have been found to be useful in ruling out major abnormalities. Examples are the biparietal diameter (BPD) plane of the fetal head, the posterior fossa, the atria of the lateral ventricles, the 4-chamber view, and the outflow tracts, umbilical insertion, sagittal, and transverse views of the uterus and endometrial cavity in gynecology, etc. Although these standard image planes can be successfully obtained in the great majority of patients, factors such as fetal position makes their visualization quite difficult if not impossible at times. The patient might have to be brought back for a follow-up scan at a later time because of nonvisualization. The use of volume ultrasound imaging significantly reduces the number of these unsuccessful instances. It also reduces the time it takes to obtain certain anatomical views that might otherwise require quite a bit of scanning and probe manipulation. These views can be generated from an acquired volume even though the probe in 2D mode could not have had direct access to those planes. In addition, volume ultrasound imaging provides access to certain image planes that are not considered standard. Typically, these nonconventional image planes are either difficult (but not impossible) to obtain, such as the midline sagittal view of the fetal brain to visualize the corpus callosum, long axis of the cerebellar vermis, and/or the brainstem (Figure 43-3), or image planes that are simply impossible to obtain. The impossible image planes are typically those that are either perpendicular or close to perpendicular to the ultrasound beam (Figure 43-4). As a result, they are physically impossible and therefore not a requirement in any 2D ultrasound protocol. These impossible image planes can be generated from an ultrasound volume, which gives the operator access to an unlimited amount of scan planes at virtually any angle. Of course, some of the main requirements and conditions of ultrasound should be met, such as acquisition plane, resolution, and depth.

These planes can be very important clinically, as they might be the best views to rule out some difficult conditions. Examples of impossible planes with clinical importance are the coronal plane of the uterus and endometrium (Figure 43-5), coronal plane of the cervical canal, coronal planes of the rectum (Figure 43-6), or the coronal plane of the perineum with short-axis views of the urethra, vaginal canal, and rectum, as well as the long axis of the puborectalis portion of the levator plate seen in one single view, to name just a few (Figure 43-7).

The second major reason to use volume sonography is that it can provide depth, which is also impossible to obtain with 2D ultrasound. The reason is simple: with 3DUS a volume of 2D ultrasound data is acquired. As a result, the beam is "thickened" and the anatomy does not have to be viewed only by cross sections via a "thin" 2D beam (see Figures 43-7I and J and 43-8). The anatomy can be analyzed in user-selectable transparent and nontransparent (surface) modes through this "thick" beam. Depth is provided for both the possible conventional planes (ie, fetal profile or frontal face view; Figures 43-9,43-10, and 43-11), and for the impossible ultrasound views previously mentioned (Figures 43-12 and 43-13). In addition, there are some impossible image views that can only be adequately visualized using depth, as their shape is not straight and visualization of different depth levels is needed. Coronal views of the perineum, fetal secondary palate, and optic chiasm are typical examples (see Figures 43-7I and J,43-12, and 43-13).

In everyday practice, either one or both of these reasons (impossible planes and/or depth) are used to gain access to certain anatomic planes that are deemed clinically useful. As a rule, we use the mutiplanar views for cross-sectional analysis using our knowledge gained over the years with 2DUS and correlate that with the many choices offered by the rendered view, which offers depth. We do this analysis sometimes in the same display, and at other times in individual displays (Figures 43-14,43-15, and 43-16).

Volume acquisition is arguably the most important part of the whole process of using 3DUS clinically.

Ultrasound volume acquisition methods are divided into 3 main categories:

1. The manual or "free-hand" acquisition method acquires volumes using a conventional 2DUS probe. Commonly, a position sensor is mounted externally to the probe assembly to provide position-sensing data to accurately register the acquired 2DUS data in space and time. The free-hand technique can be used in both in-line (where image acquisition, transfer, storage, and manipulation is done in the same unit) and off-line systems. With off-line systems, volume data are commonly captured using the ultrasound machine's video out port and downloaded to an off-line computer workstation equipped with a video frame grabber PCI card. Volume imaging software is used for subsequent 3DUS image reconstruction and analysis. Although the free-hand technique can be economically implemented into existing systems as it uses conventional 2DUS probes, its main drawback is that it cannot be used for accurate measurements, and its volumes are likely to suffer from acquisition-related artifacts.^7,25 With off-line systems, the external probe-holding mechanisms are typically bulky and the image transfer process to an off-line workstation is labor intensive and time-consuming (Table 43-1). The free-hand acquisition techniques are traditionally used to image parts of the anatomy that do not move (the nonpregnant uterus and adnexa) and generally are avoided when imaging moving targets such as the fetus (Figures 43-17 and 43-18).

2. Mechanically swept 3DUS probes (ie, probes with a mechanized drive contained within the probe case itself) are commonly referred to as dedicated volume probes. When these probes are activated, the transducer elements are swept by a motor through the region of interest within the limits of a volume box, with an operator-selected speed and angle, while the probe is held stationary. The most common design of these probes has an electronic array achieving imaging in one direction (the 2D beam) that is almost simultaneously swept mechanically in the orthogonal direction (volume acquisition in the Z plane). So, 2 orthogonal planes are acquired: one is achieved electronically, and the other mechanically by sweeping the beam. The main advantage of this method is that images obtained with the automated approach permit accurate 2D and volume measurements and demonstrate more precise spatial relationships. The dedicated volume acquisition methods are considered more reliable, especially in imaging moving targets. As the operator's hand stands stationary and the transducer elements move through a motor, the volumes can be acquired and displayed fast, up to 42 times per second (at the time of this writing), allowing the depiction of moving targets as they change position (fetus, fetal heart). This fast acquisition and display method is commonly called four-dimensional ultrasound (4DUS) with the element of time considered the temporal or the fourth dimension (see Figures 43-17 and 43-18). One major drawback of this method is that it does not focus in the elevational plane, or the plane along the depth or volume acquisition axis (see Table 43-1). While the volumes can be also acquired with color or power Doppler, as well as with B-flow, these acquisitions cannot be obtained fast and in continuity in a 4D sequence. They can only be obtained as one single volume, and the volume acquisition time is typically longer compared with gray-scale-only volumes.

3. Two-dimensional matrix arrays. These transducers represent the future of volume acquisition. They are composed of a transducer mosaic arranged in rows and columns with a very high number of 2D piezoelectric elements (4000 to 8000 elements and even more projected in the near future). These probes scan in both orthogonal planes electronically and can acquire a pyramidal-shaped volume.^{14,26,27, and 28} Because these transducers allow volume acquisitions in all planes they allow focusing not only in one plane, but in the elevational plane as well, significantly increasing the quality of the acquired volumes. In addition, these transducers provide very fast frame rates without significant deterioration in image quality because all the transducer elements are fired at the same time (or in sequential groups). To increase the area of interest, they may employ 3D beam steering (see Table 43-1). The information is typically displayed in a biplanar display or typical orthogonal multiplanar displays, as well as in rendered displays. Due to the huge amount of data these transducers collect, the hardware and software needed to process this information is still in its early days and will need to be improved (see Figures 43-17 and 43-18). One workaround for this problem for the moment is the reduction in the number of available channels, otherwise called sparse arrays, which results in inferior image quality but the data can be processed faster.

The 2D matrix arrays can acquire the following: (1) A single 3D volume, which is typically acquired very fast, in roughly 1 to 2 seconds; but again, the speed of acquisition depends on volume size. (2) A 4D volume sequence, where the volumes are acquired one after another, but they are not that fast because all of the elements are sequentially fired at full aperture, so they do not yet acquire in real time. These volumes can also be acquired with color or power Doppler and this is one of the strongest points for this mode. (3) Fast 4D real-time imaging of up to 24 volumes per second. The trade-off here is that the transducer only uses roughly one-third of the aperture. (4) Real-time biplanar imaging where 2 orthogonal images are acquired and displayed in real time in gray scale as well as with color or power Doppler.²⁸ The biplanar image is fully focused and can rotate as well as tilt laterally in real time. This particular mode is extremely helpful and has a central role in the use of 2D matrix arrays.

Matrix arrays are the next phase in volume acquisition. Their main advantages are related to the fast acquisition speeds, significantly reducing motion artifacts (especially useful in imaging the fetal heart), and focusing achieved at 360 degrees, making possible visualization in any image plane with no loss in resolution. Their narrow apertures and still relatively low-frequency probes are their main drawbacks at the moment.

Gray-Scale Scanning and Acquisition Guidelines

Before a volume is acquired, certain important principles should be considered. Considering that the most widely available and used probes currently in the market are the mechanically swept ones, these guidelines deal mainly with issues pertaining them, even though many of these principles apply to matrix arrays as well.

Prior to volume acquisition, optimize the 2D image to maximize the quality of the volume by adjusting depth (improving the refresh rate and making the acquisition faster), aperture (or beam angle, improving lateral resolution), number of focal zones (one is usually enough), and harmonics. If imaging the fetal heart, lower or even completely remove the persistence value, as frame averaging does not adequately display the valve motion. Pay close attention to the refresh rate value displayed in the screen, as a lot of the preprocessing options such as compound resolution imaging, speckle reduction, or coded excitation may negatively influence it (Figure 43-19).

Acquisition of volumes can be accomplished via both transabdominal and transvaginal transducers. In gynecology and first-trimester acquisitions, although the quality of the transabdominal acquisitions and rendering is adequate, transvaginally the quality of the volumes is generally very high. Otherwise, second- and third-trimester acquisitions are performed transabdominally. The examination is started in conventional 2D mode to target the region of interest (ie, the fetus), preferably in a long-axis view. The aperture is kept at the narrowest possible but still including the longest axis of the fetus. Acquisition is started with the lowest speed that translates into the highest quality setting. As a rule, the faster the acquisition speed, the lower the number of 2D slices acquired in the volume. Therefore, the multiplanar analysis in the planes other than the acquisition plane is likely to display low-resolution quality. If the fetus is moving, a fast 4D acquisition is needed to keep up with the fetal motion and avoid artifacts. The slower the acquisition speed, the greater the number of 2D slices in the acquired volume. Consequently, slower acquisition speeds yield higher resolution volumes. On the downside, lower speeds typically are more likely to suffer from motion artifacts (especially when imaging a moving target) (Figure 43-20). Registration artifacts are more prevalent in the views other than the acquisition plane.

It is important to maintain an even pressure with the transducer over the area of interest. Otherwise, distortions from uneven pressures can be seen on superficial (near-field) structures. Another important aspect is maternal breathing. If politely asked, most patients will have no problem holding still for 5 to 6, or even up to 10 seconds. The operator too should be still (in automated scans) and maintain good contact with the patient's abdomen during acquisition. If scanning with a free-hand transducer, maintain a constant acquisition speed as you move over the area of interest with your hand and try to keep the acquisition planes parallel and equidistant from each other.¹

At times, the acquisition plane of interest is simply not in a favorable position. Most of the time, the spine can be brought down (or up, if the spine is of interest) by reorienting the probe typically from the side where there is some fluid. An effort should always be made to use a pocket of fluid between the probe and the acquisition plane. If the fetal face is right up against the placenta and there is no fluid interface between them, slightly releasing the probe might help bring some fluid in. If not, another way to deal with this is to gently push either the frontal part or the chin of the fetus with your left hand until some fluid comes in. The best rendered volumes will come from good tissue interfaces that were acquired by rigorous scanning.

At the end of the acquisition, the volumes are explored for the presence of artifacts, and if artifacts are present the volume is discarded and a new one is obtained. If artifact-free, it is saved in the hard drive of the machine as a 3D volume, or if a 4D sequence is acquired, as a 4D sequence, a 3D volume, an .avi video file, or all of the above.

Acquisition plane matters (see Figures 43-3,43-21,43-22,43-23, and 43-24). Certain diagnostic planes can only be obtained within a volume acquired in a specific acquisition plane that is proven to generate the expected image plane.^13,29 A good practice is to always scan in 2D ultrasound over the area where volume acquisition will take place. With practice, the operator will know what volumes will or will not generate the desired views just from scanning in 2D over the area of interest. Within the 3D volume, resolution is highest in the plane of acquisition or close to it (for example, sagittal planes through the uterus in a volume acquired with the probe in the sagittal orientation). Resolution is progressively degraded as the image plane deviates from the acquisition plane (see Figure 43-24).³⁰

Examples are acquiring the fetal heart volumes at the level of the 4-chamber view with the fetal spine below the 9 or 3 o'clock level yields the highest number of heart views. Fetal spine volumes are generally best acquired when the spine is above 10 or 2 o'clock. The closer the acquisition plane to the anatomic plane desired, the better the chance that it will be generated from the volume. Acquisition planes close to 90 degrees to the desired planes are generally the poorest in resolution. As a result, the coronal reconstructed plane generally has the lowest resolution of the 3 planes in the multiplanar display. Manufacturers commonly interpolate (as in fill in the empty spaces to make the image more understandable) the C plane (the plane 90 degrees to the axis of the beam) to improve the quality to a certain degree.

The concept of acquisition speed is often misunderstood. The misconception has been that the faster the acquisition speed, the better that particular system should perform overall. Add to this the marketing hype of new systems across the industry with terms such as 3D live, 3D real time, 4D, 4D live, and 4D real time and you can see why the casual user may get the impression that going from live to real time and/or from 3D to 4D (adding more Ds) is what it takes to improve image quality. There are quite a few 3DUS users that routinely use 4D acquisition as the default acquisition method whenever a volume is contemplated. With mechanized volume probes, which are the main probes used for 3DUS today, you mostly aim to acquire one volume at a time, with the slowest speed for best quality. If the target is moving, either get a faster 3D volume or use 4D mode, which acquires several volumes one after another, and on the fly it displays them on the screen at the selected display mode (see Figure 43-19). In short, you either get one 3D volume or many 3D volumes in faster speeds. How fast should you go in 4D mode? The slowest possible (with no artifacts) the better, is the answer!

What about real-time 4D ultrasonography (4DUS)? The real-time concept needs to be explained. By film and graphics industry criteria, 30-frames-per-second update of any image in a TV monitor (NTSC video in North America) is generally considered real time. That does not mean below the 30 frames per second the video speed is slow. Film has been traditionally shot and displayed at 24 frames per second, and PAL-SECAM video in other parts of the world is also updated 24 frames per second with equal clarity to NTSC. But as a general rule, over 30-volumes-per-second acquisition and over 30-frames-per-second display are considered real time. If you keep the volume box really thin, most equipment today can achieve acquisitions over 30 volumes per second. In clinical practice, this concept takes a different value. What we consider important is sampling motion fast enough to see its different points in time (temporal resolution) by maintaining acceptable anatomical detail (spatial resolution). These variables differ among structures. Motion of the extremities can be abrupt, but usually they are of slow speed. So the sampling rate does not have to be high, and usually 5 to 10 volumes per second are sufficient to visualize them adequately. On the other hand, fetal heart chambers and valves move quite fast and a higher sampling rate would be required, something only 2D matrix arrays can really offer. For the moment, we use a dedicated program for the fetal heart called spatio-temporal image correlation (STIC). Commonly, we scan fetal extremities with up to 5 to 10 volumes per second (sometimes a bit faster or slower depending on volume box size) and do not think much about the concept of "real-time imaging" rather than opening up the volume box enough to include the whole area and visualize motion. Volume depth directly affects acquisition speed, and in order to get over 24 volumes per second we would have to make the volume width thin and reduce depth. That would be impractical with current equipment, as most fetal structures would need a reasonable volume box width and depth.

One example of a useful display technique is volume contrast imaging (VCI; GE Healthcare, Milwaukee, WI, USA) (see Figure 43-8). The VCI technique can be a static, 3DUS-based, post-processing function, or a 4D ultrasound–based function consisting of several layers of 2D slices acquired fast in continuity and displayed as a volume-rendered view. Volume-rendered view collects tissue signals from adjacent deeper layers and project them as a conventional 2D view either in the same view as the acquisition plane (VCI-A) or in the plane perpendicular to it (VCI-C). In both those planes it allows standard 2D measurements. The slice thickness is adjustable by the user, ranging from 2 to 15 mm. The surface-rendering mode uses a proprietary algorithm.³¹

VCI-C displays a coronal slice as a 2D plane with the extra advantage of better tissue contrast compared to conventional 2D. The proprietary surface algorithm separates the noise from the voxel data and subtracts it. As a result, contrast resolution is improved. Gray-scale information is projected into a 2D plane provided by the rendering of all data within the thickness selected. A mixture of 2 render modes is applied to the data: (1) surface mode, which retains both the original gray levels of the tissue as well as image sharpness, and (2) transparent maximum mode, which contributes tissue contrast derived from deeper layers of voxels. The effect of removing noise and enclosing data from adjacent layers results in a smoother image combined with enhanced contrast resolution. The reason this algorithm was developed was to improve the detection of tissue interfaces that normally are not appreciated with 2D ultrasound. In everyday practice, it is a very straightforward method and can be used similarly to a 2D beam with the added benefit of better contrast resolution and less noise.³²

3D/4D Color/Power Doppler Sonography, B-Flow, and STIC Acquisition Guidelines

All of the described methods involve volume acquisitions that incorporate color and power Doppler ultrasonography, as well as B-flow (Figures 43-25,43-26, and 43-27). It is important to remember that 4D acquisitions with mechanically steered probes cannot acquire color Doppler, power Doppler, and B-flow. With these probes, STIC can display what appears to be a 4D ultrasound display, but it really is not a 4D acquisition in itself. Matrix arrays can obtain biplanar, multiplanar, and rendered real-time acquisitions with color and power Doppler and B-flow of very high quality. Knowledge and optimization of the settings are the most common problems ultrasound operators encounter in their learning curve of acquiring 3D color and power Doppler volumes. There are no fixed recipes for every occasion, but there is a reliable set of principles and requisites that, if understood, should be helpful in obtaining adequate information. Normal 2DUS principles of visualizing vessels with color and power Doppler assist in a 3D volume acquisition as well. Here is a listing of the most important concepts:

1. The scale or pulse repetition frequency (PRF) should be adjusted according to the particular blood flow speed and vessel size. A higher PRF is typically needed for higher velocity vessels (ie, intracardiac flow), and a lower PRF is needed for smaller and/or branching vessels (ie, brain) especially placental or retroplacental (venous) vessels.

2. The operator should orient the 2D beam to be as parallel to the vessel or vessels of interest as possible.

3. The overall gain should be adjusted to avoid "bleeding artifacts" in the vessel wall by decreasing its value at the minimally acceptable level, typically no more than 50% to 70%.

4. The color or power box should fit tightly around the area of interest. The color or power mode 2D frame size does count in relation to frame rate.

5. When scanning in 2D, try to go over the area of interest with color or power Doppler and see whether you can get any signal. If not, it is likely you do not have the right angle to the vessels of interest (parallel or close to it). As a result, prior to acquiring a volume with color and power Doppler, scan in real time and adjust the angle first and make sure you get signal.

6. The wall filter should also be adjusted. The medium settings appear to work better most of the time.

7. Color priority set on high generally displays better quality, and it makes sense to use it whenever color or power Doppler ultrasonography is contemplated. If available in your equipment, set the dynamic motion differentiation to avoid background noise signals.

8. Pay attention to the sensitivity settings. They should not always be put at high, unless you are looking at the retroplacental vessels. The medium levels usually work best in most scenarios.

A good practice is to start in 2D gray-scale imaging with the highest frame rate achievable for that particular frequency and penetration. To do so, the user needs to understand the variables to be controlled that influence the frame rate. A high frame rate translates into a volume with higher quality, as there will be more frames in a volume. The most important factors affecting the frame rate are as follows:

1. The aperture (or angle) opening in the 2D beam (use the narrowest angle you can).

2. The number of focal zones, and the amount of depth (the less the better for both).

3. Harmonic imaging helps significantly in imaging in the near and mid-fields but not much in the far field. It is important to know that it may slow down the frame rate.

4. Try to put the area to be acquired in the near field of the transducer. If not possible, at least see if you could put a good pocket of fluid in between. This will ensure a low attenuation acoustic window and improve penetration and resolution.

Acquiring Volumes with B-Flow

B-flow or B-mode (brightness mode) is a relatively recent technology that displays the reflected echo intensities as bright dots. A subtraction mode is used to differentiate and visualize the "bright" amplitude of scattered particles. The unique feature of B-flow is that it can visualize the cardiovascular system with a technology that does not depend on the frequency shift. B-flow is based in gray scale, and that it does not depend on the frequency shift means that there is no angle dependency. It also does not display any bleeding artifacts or aliasing commonly seen with color and power Doppler sonography. This algorithm can display the direction of flow but not as well as color Doppler technique (Table 43-2). The resulting image in a rendered volume is a view that displays the cardiovascular system like an angiography but in a volume with depth, so it can be rotated 360 degrees. These B-flow volumes can be acquired as a single 3D volume, as a 4D sequence, and also as STIC volumes.³³ They are best rendered with maximum intensity projection, but a combination with surface or light modes can be useful as well (see Figure 43-26B).

Spatio-temporal Image Correlation

In the past it was believed that increases in acquisition speeds will have a great impact in 3D fetal echocardiography. Although mechanized volume probes can achieve speeds of acquisition of over 40 frames per second, acquiring fetal heart volumes at these high speeds does not necessarily translate into higher quality volumes. To solve these issues, a new volume acquisition technology was introduced called spatio-temporal image correlation (STIC). With STIC a very slow (always by 3DUS standards) acquisition from 7.5 to 15 seconds is performed over a preselected area of the fetal heart, typically at the level of the 4-chamber view.^11,34 The acquisition angle varies from 15 to 40 degrees and is user selectable.³⁵ The same image optimization criteria apply as the ones mentioned above. After the acquisition, post-processing of spatial and temporal data is performed so the 2D-acquired images are correlated in time and space. The acquired data are displayed in a classic multiplanar view and/or in a cine sequence depicting heart motion with total control on interactive re-slicing and/or rendering the same as in a static 3D volume.^{11,16,34,35,36,37,38, and 39} Typically, with gray scale, this acquisition has a very high B-mode frame rate (approximately 150 frames per second) due to the relatively small region of interest.

STIC was initially introduced only in gray scale, but later the acquisition was achieved using color and/or power Doppler technology as well. This development opened up a whole new area in evaluating the fetal heart with 3DUS. STIC represents a significant development in the field of volume fetal heart scanning having the potential to perform a full fetal echo exam out of a single fetal heart volume at the level of the 4-chamber view. Studies done prior to and after the introduction of STIC were able to demonstrate that a volume of the fetal heart can provide all the standard imaging views for a fetal echocardiography (see Figures 43-27 and 43-28).^{16,39,40, and 41}

Although the potential of this technique is clear, there are a number of pitfalls worth mentioning. The acquisition is still relatively slow, from 7.5 to 15 seconds, and as such fetal breathing and gross body movement can occur. Careful evaluation of windows B and C in multiplanar display after the acquisition is crucial to ensure that no artifacts are present. Adjusting the volume box tightly around the region of interest helps as it generally increases the frame rate. The best frame rates recommended are the ones above 15 Hz, typically with higher rates in smaller hearts and slower rates in bigger hearts. An important adjustment is the acquisition angle, as the default settings tend to be very narrow. Optimal angles vary from 15 to 35 degrees and are related to the gestational age. As a rule, the earlier the gestational age the narrower the angle. STIC acquisitions perform well in spine-down fetal positions, but not so well, if at all, when the fetal spine is up. Ideally the fetal spine should be maintained below 9 o'clock to 3 o'clock with the apex of the heart pointing up. As a rule, apical 4-chamber views are the preferred acquisition planes, while the lateral and basal 4-chamber views are inferior or even unusable at times. While axial views of the chest at the 4-chamber view are the best overall acquisition planes with STIC, ductal/aortic arches and inferior vena cava (IVC)/superior vena cava (SVC) can be best extracted from sagittal acquisition planes with the currently used mechanically swept transducers.⁴² With 2D matrix array probes, this will not be the case as they focus in both directions. It is important to understand that while STIC is a useful technology for the moment, it will eventually be replaced by 2D matrix arrays, which improve precisely where STIC suffers most: elevational focusing, slow frame acquisition times, and motion artifacts. Last but not least, STIC remains a great tool to learn fetal echocardiography.

The saved volume data can be navigated throughout to demonstrate innumerable arbitrary planes. The multiplanar display is the most versatile and important volume display, where any anatomic structure is interactively analyzed in 3 orthogonal planes displayed simultaneously at the level of the cross planar point. Most manufacturers label them as windows A, B, and C, representing the X, Y, and Z planes, although as you manipulate the planes this representation may change. This is the most important display to learn and use. Right after acquisition, window A represents the acquisition plane. The volume can be explored as desired by scrolling through parallel planes, just like in axial CT views, in any of the 3 views and/or by rotating the volume in order to obtain an optimal view of the structures of interest. Typically there are X, Y, and Z knobs that rotate these windows in 3 directions, as well as a parallel shift knob for axial parallel moves. An unlimited number of multiplanar displays are possible, including variable degrees of magnification. These 3 planes are interactively related to each other orthogonally and this relationship is always maintained, even after reformatting. Correlation between these 3 planes is used to confirm a given desired plane, such as the midsagittal, midtransverse, or midcoronal plane. This correlation in the multiplanar display is accomplished by placing the planar center point at the point of interest in one of the planes and observing the location of the corresponding center points in the other 2 planes (see Figures 43-15 and 43-16). This is arguably the most powerful feature of the multiplanar display and the most used one. For example, the exact location of a small hypoechoic area initially noted in one view can be confirmed by placing the planar center point in the middle of that area, then identifying that the point is also in the middle of a small hypoechoic mass in the other 2 perpendicular planes. In addition, all or part of the saved volume can be processed and displayed into a rendered image that can be displayed alone or in correlation with the multiplanar display. Various similar interfaces are developed from different manufacturers, and while they may have interface differences, all the main principles highlighted above apply (see Figures 43-3,43-6,43-7,43-12,43-13,43-14,43-15, and 43-16,43-29, and 43-31). New forms of image display that enhance the multiplanar display are being introduced, improving the ability to demonstrate complex anatomic relationships. In obstetrics and gynecologic applications, the most commonly used setup is the classic multiplanar display, which can be simultaneously viewed and correlated with the rendered display as well (see Figures 43-3,43-12,43-13,43-14,43-15, and 43-16, and 43-26). New tomographic imaging displays are emerging where the user has the freedom to select any curved scanning plane around any irregular surface (versus straight) (see Figures 43-5,43-12,43-13,43-28, and 43-31). These planes are not necessarily parallel to each other, but the user has the freedom to change the scan plane axis according to need. The operator has the ability to bend the scanning plane to conform to the anatomy, making these displays very versatile and allowing the user to evaluate the anatomy not only in straight lines (Figures 43-12,43-13,43-28, and 43-31). Most of the time data manipulation is done right after the acquisition very quickly to see that the desired view has been captured, ruling out any anomalies. For more complex scenarios, data manipulation is done after the clinical examination has been completed and the patient discharged.

An important development has been the description of standard techniques for navigating volumes of complex anatomical structures and extracting the required views. As an example, knowing the spatial relationships between heart views, a step-by-step technique could be designed utilizing the orthogonal relationships of the multiplanar display around the cross-planar point. Several similar techniques have been described for the fetal heart and even the uterus.^{42,43,44,45,46,47,48, and 49} Of course the extraction of these views can be automated. Abuhamad et al went a step further and tried to quantify the relationships of the heart views stratified by gestational age and fed this information to an algorithm that tries to retrieve the views out of a tomographic ultrasound image display called SONOVCAD (GE Healthcare, Milwaukee, WI, USA).⁵⁰ This is a great first step forward in automating scan plane retrieval that opens up new applications as well. The problem is that this algorithm at the moment still has to somehow "obey" the strict orthogonal limitations of the current tomographic displays, and is not flexible enough to allow unrestricted moving of the cross-planar point—a key step in image plane retrieval. Newer, more flexible tomographic displays should make this already very useful tool even more versatile (see Figure 43-28).

In digital imaging, the smallest unit of an image is the pixel. The smallest image unit in an ultrasound volume is a voxel, or a volume pixel or pixel with depth (Figure 43-32). The qualitative analysis of volume data involves adjustments of texture, brightness, and transparency of voxels (Figure 43-32). Typically these adjustments are performed via ray casting (or ray tracing) algorithms mimicking the physics of light reflection via computer-generated graphics. With the ray-casting approach, a number of rays are projected from the 3D volume to the viewer's eye. Volume data are optically transparent, and variables such as transparency, shading, and reflectivity are dependent on the intensity of echoes inside the volume as viewed depending on the algorithm selected. Usually, each voxel is characterized by 2 different values: an opacity value that defines the light quantity that crosses the voxel, and a shading value that represents the reflection or diffusion properties of the tissue. Shading and opacity/transparency voxel values along each ray are typically examined, multiplied by factors, and summed to achieve the desired rendering result. A wide spectrum of visual effects can be generated depending on how the algorithm interacts with each voxel encountered by a particular ray. At the end of the acquisition of a number of 2D images, the data are converted into a 3D volume by 3 main reconstruction algorithms: (1) pixel based, (2) voxel based, and (3) function based. The most common approach is the voxel-based one where each 2D image is put at the proper location in the 3D volume. Using ray-casting algorithms, one or several pixels may contribute to the value of each voxel. This approach preserves all the original information during 3D reconstruction and allows evaluation of the 3D volume by a variety of rendering techniques. Using a voxel-based volume model, the operator can scan through the data and then choose the most suitable rendering technique. Moreover, this approach allows the use of segmentation and classification algorithms to measure volume and segment boundaries, or various volume-based rendering operations.⁵¹ Segmentation is the process of extracting surface or at-depth data from the acquired volumes. A number of different semiautomatic and fully automatic segmentation methods have been developed.⁵¹

Volume data are visualized with 2 main techniques:

1. Surface rendering

2. Volume rendering

In surface rendering the algorithm highlights surface tissues (typically soft tissues). Surface rendering can be processed manually or automatically, but the manual methods predominate at the moment. Brightness, texturing, and shading/lighting are used to improve depth perception, and increasingly operators are given a multitude of tools to adjust their values (Figures 43-33,43-34,43-35,43-36, and 43-37). Depth perception is typically improved by different computer-generated shading and lighting effects where the surface perpendicular to the ray reflects higher intensity values compared to the area at an oblique angle with the rays.

Volume-rendering methods display transparency by mapping all the voxels in a volume directly onto the screen. They require that the entire data set be sampled each time an image is rendered. Maximum and minimum intensity projection (MIP) methods are the most useful volume-rendering options (see Figures 43-34,43-35,43-36, and 43-37). They are a form of ray casting where only the maximum (or minimum) voxel value is retained as the rays transverse the data volume. Volume-rendering algorithms typically serve to display 2 main areas: (1) the fluid-filled or hypoechoic areas (ie, minimum intensity projection or inverse mode) and (2) the bony tissues (maximum intensity projection). The inverse mode is a form of volume rendering where any hypoechoic (low-intensity) voxels are tagged and inverted to look hyperechoic, and all hyperechoic voxels are tagged and inverted to look hypoechoic. The end result is a rendered view of all the cystic portions within a volume, making the solid areas transparent.⁵² This mode assists the user in better defining spatial relationships between vessels and/or between cystic structures. The rest of the rendering options are typically intermediate ones and less useful. They mainly deal with lighting and mid-intensity projections.

Up to 2 user-selectable rendering algorithms can be applied and mixed together, each contributing a certain percentage to the end result. Typically, surface and smooth or surface and light settings are combined with each other, along with adjustments of the gamma curve and lowering the threshold for optimal results. Gamma curve is a histogram function very useful in adjusting brightness and contrast of any volume. The most commonly used gamma curve adjustment is in a graphical display by changing its flat shape to an S shape to various degrees with the user visually controlling the results. Another typical manipulation with both surface and volume rendering is the threshold. There are 2 thresholds: the low and the higher threshold. The low threshold removes low-level (intensity) echoes and is the most commonly used one. The high threshold removes high-intensity echoes (like bone) and typically is not used as much as it may remove too much useful information. Some manufacturers do not even give the user access to it. Last but not least, the transparency level of a volume can also be adjusted and, depending on the user's goal, can be lowered or increased. It is usually expressed as a percentage with 50% typically being the desired value.

Volume data can be edited via a number of editing tools. The most common one is via the editing boxes where the user narrows the displayed volume box either in straight lines or in a curved fashion along with the use of low threshold. Another very useful editing tool is the electronic scalpel where the user selects and removes part of the anatomy that is superimposed to the area of interest. The editing tools work best whenever there are different and well-delineated tissue interfaces present, such as soft tissue versus fluid.

With 3DUS, for the first time ultrasound-based methods allow very accurate volume measurements, even of irregularly shaped structures. Quantification analysis of volume data requires some very precise segmentation and selection tools. Typically, ultrasound machines do not yet offer precise input devices to trace organ systems but commonly expect the user to use the trackball. Fortunately, most of the volume ultrasound manufacturers today offer software packages to further manipulate and post-process volume data sets off line in a desktop. Some examples include fourSight View–Tool by Siemens Medical Solutions, SonoView Pro by Medison, 4D View by GE Healthcare, Q-Lab by Philips Medical Systems, and Sonocubic MyLab by Biosound Esaote. These programs offer an equal number of features and controls commonly available in the systems themselves, and at times may offer even additional post-processing, selection, quantification, and archive tools. In the desktop, the user can make use of available third-party input devices such as a digitizing pen or a high-precision laser mouse for finer control in tracing.

The volume-derived quantitative tools can be best described in 3 main categories as follows:

1. Extraction of standard anatomical planes and quantification of their relationship with each other

2. Volume quantification of organ systems

3. Blood flow quantification and tissue perfusion studies

Extraction of Standard Anatomical Planes and Quantification of Their Relationship With Each Other

Commonly in clinical practice there is a need to obtain objective analysis of the ultrasound findings. This process can be very difficult with 2DUS, as this analysis might need to be performed in more than one image plane. Distance and volume measurements from an adequate volume ultrasound have been proven to be accurate.⁵³ Volume ultrasound is very useful in standardizing image planes and quantifying relationships. A very common application is the precise identification of the midsagittal or midcoronal planes of any anatomic structure such as the uterus,⁴⁵ cervix,⁵⁴ face profile, nasal bone,⁵⁵ phallus,⁵⁶ brain,⁵⁷ etc. Techniques have been developed for standardizing the extraction and objective evaluation of important planes such as fetal iliac angle measurements,⁵⁸ a diagnostic approach for the evaluation of spina bifida (see Figure 43-30),⁵⁹ a 3D sonographic approach to the diagnosis of retrognathia and micrognathia,⁶⁰ extraction of the coronal plane of the uterus,⁴⁵ manual and automatic extraction of the heart views,^{42,45,46,49,61} etc. These techniques are very useful in daily practice and worth studying for anyone contemplating to use 3DUS clinically.

Volume Quantification of Organ Systems

There are 2 methods to trace an area in order to measure volumes. One is the parallel tracing method where the operator stays parallel to the acquisition plane. The other is the rotational method where volume calculation is accomplished by equally dividing the 360 degrees of the volume of interest in slices of 6, 9, 15, or 30 degrees along a centrally fixed rotational axis. Both methods are very reliable and reproducible. The parallel tracing method usually is more accurate in identifying landmarks and tissue interfaces as it stays in the acquisition plane, but it has a problem with the consistency of the number and distance of the traced slices. A commonly used program for measuring volumes is the Virtual Organ Computer-Aided Analysis (VOCAL) program (GE Medical Systems). The volumes are measured by tracing the area of interest starting at the region of interest, and the next slice is rotated at a predefined angle. The measured volumes can be displayed as color-coded/shaded or wire-framed models that are representative of volume data (Figure 43-38). Although the rotational method seems to impart some consistency and standardization in the measurements, the tracing plane shifts from the acquisition plane as the operator proceeds. Shifting from the acquisition plane makes the tracing of the organ (ie, fetal lungs) very difficult because the tissue interfaces may be lost the farther you go away from the acquisition plane, and the operator might have to estimate the border of an organ.

One study showed that the rotational (VOCAL) volume measurement method is associated with higher interobserver variability and a lower level of agreement when compared with the parallel tracing method.⁶² In another study, Raine-Fenning et al evaluated the interobserver reliability and validity of measurements of phantom objects of known volume using conventional and rotational techniques of volume calculation according to the measurement technique.⁶³ They found no significant differences in reliability on the basis of the measurement plane or between observers, but this was a phantom study where the interfaces can be easier to delineate.

Several technical factors have limited the wide applicability of volume assessment. The major drawback is the significant amount of time required to perform these measurements. Typically, the operator is required to manually draw circumferences around the area of interest, and this process may take a relatively long time (by ultrasound measurement standards), from 3 to 5 minutes all the way to 10 minutes depending on the complexity of the structure and operator's experience. Another important and common technical problem is the presence of shadowing at the level of the distal portion of humerus or femur (when measuring arm and thigh volumes to use as a better predictor of fetal weight), making the visualization of soft tissue borders near the hip joint or the knee very difficult if not impossible at times. Some of these problems could be reduced by the application of the fractional limb volume measurement, where the operator only measures the volume of the middle aspect of the limb where the ultrasound tissue interfaces and the borders of the soft tissue are generally easier to depict.⁶⁴ The midportion of both upper arm and thigh is identified in the multiplanar display, and the circumferences are drawn in the perpendicular plane in orthogonal views.

In daily practice, volume ultrasound measurement tracings and the volume itself, in its reconstructed and perimeter-drawn state, can be saved as such. In other words, volumes can be saved in the hard drive of the machine or an external one with the calculated volume in place with some software (ie, VOCAL). This largely unknown ability of the VOCAL program comes in very handy in assessing the quality of the volume measurements. This is an important step that, from our experience, serves well for training purposes, as well as for audit and/or quality control. The saved volume as such provides detailed evidence of how the operator arrived at such a value. What is even more important, these volumes can be not only analyzed, but also redrawn and requantified, if the examiner finds it necessary.

There have been a great number of studies estimating volumes of organ systems and generating nomograms across gestation as well as in gynecology.^{20,21,23,64,65,66,67, and 68} Most of them do seem to prefer the VOCAL-based technique (see Figure 43-38). Nevertheless, the reader should understand that currently the use and integration of organ volume measurement in clinical diagnosis and management is still limited.

Recently, Deter et al presented a very original paper describing the development of a quantitative method for characterizing the gestational sac shape. This is the first paper that shows the potential for a new era in volume ultrasound by providing quantitative analysis for any biological shape and its changes over time in normal and abnormal processes.⁶⁹

Blood Flow Quantification and Tissue Perfusion Studies

The goal of obtaining volume data from power and color Doppler ultrasonography is not only to visualize the vascular tree of an organ system, but ideally to also give an overall estimation of the blood supply of a given anatomical area. Efforts have been made to further post-process this information with the purpose of quantifying the amount of blood supply. A commonly used program is the 3D-Shell. With this method, a 3D volume is acquired with power Doppler. A volume of interest (VOI) inside the initial acquired volume is identified and traced. Inside the volume of interest the gray-scale and color histograms can be evaluated together or separately and expressed as a mean value. The expected clinical value of these power and color Doppler volume quantification methods is better estimation of blood flow within the area of interest (ie, lung or tumor). The flow analysis is expressed with 4 indices: (1) average gray value of noncolor voxels (MG); (2) vascularization index (VI): the ratio of the number of color voxels to the total number of voxels inside the VOI; (3) flow index (FI): the ratio of the sum of color intensities to the number of color voxels inside the VOI; (4) the ratio of the sum of color intensities to the total number of voxels outside the VOI (VFI). Although these indices provide an interesting method in quantifying blood flow and vascularity, their information is based on digitized color voxel information. Power Doppler measures the amplitude of wave scattering, which depends on many variables such as sensitivity, gain, penetration, etc. The values generated may change (ie, be greater) with a more sensitive transducer, software updates, or different platforms. Published nomograms for vascularity cannot be used for imaging with other types of machines or with different generations of the same probe because of variable color sensitivity for different depths. Also, the concept of blood flow is not static but dynamic. A better way to express blood flow is unit of blood for unit of time and the above indices are static. Perfusion is a concept that involves unit of blood in a unit of time in a unit of volume. There is a critical need for deeper understanding and standardization in this area. Until then, there is no point in performing and publishing nonstandardized studies.^70,71 This topic is further examined in Chapter 47.

Volume ultrasound imaging opens up a new era in ultrasound documentation, storage, and networking. With conventional 2D imaging, the ultrasound scan has been traditionally documented via a hard copy printer or analog video recording. Most of the newer 2D ultrasound systems offer digital storage capabilities where 2D images are saved in the built-in hard drive of the machine as a BMP, TIFF, or JPEG file format. Few others offer the capabilities to record digital video sequences in the hard drive of the machine or in an externally attached digital video recorder. Volume ultrasound data are typically stored digitally. They can be stored as a single raw 3D volume, as a processed 3D volume, or as a 4D sequence, which typically saves a number of consecutive volumes. The 4D sequences can also be saved as 30-frame-per-second AVI video files. All of these volumes can be saved in the hard drive of the machine or in an external hard drive or even in a PACS system, provided they accept that particular manufacturer's file format. The saved files can be accessed from the machine via magneto-optical disks, compact disks (CDs), increasingly now on DVD disks, and recently even via ubiquitous USB flash drives. Prior to transfer, to save space, volumes can be compressed with lossless algorithms and even be encrypted for security if they are to be sent over networks.

A significant benefit of 3DUS lies in the areas of documentation, storage, and networking. Digitally saved volumes of patient data can be readily transferred to a remote site for interpretation or second opinion consultation. The implication is that primary clinical sites in remote areas would have access to expert consultation without physically traveling to the specialist's office.

Artifacts are a major issue with 3DUS, as they are with conventional 2DUS. Their sources and manifestations need to be understood in order to avoid misdiagnoses. With 3DUS, the same types of acoustic artifacts that occur with conventional imaging are encountered, such as acoustic shadowing, enhancement, refraction, reverberation, and motion artifacts from gross body movement, fetal and maternal breathing, etc. With 3DUS, there are also artifacts that are specific to the volume acquisition and display process (Figure 43-39). These artifacts have been divided into acquisition, rendering, and editing artifacts, as discussed in detail in a review article by Nelson et al.⁷² Motion-related artifacts are very common in obstetrical scanning (fetal breathing, gross body motion), but can also be encountered in gynecology (from pulsations of the abdominal aorta or the patient's breathing).

Most commonly, artifacts are seen during long acquisitions such as STIC or volumes acquired with color or power Doppler. In shorter acquisition times, asking the patient to hold still for a couple of seconds usually eliminates the maternal breathing issue. Fetal motion can be a challenge at times, but very commonly fetal movement is seen also as an opportunity to gain access to difficult planes. The bigger problem usually is a fetus that does not move much and is "stuck" in an unfavorable position. In imaging the uterus and the cervix, one of the most commonly seen artifacts has been termed an enhancement artifact in which the echogenic endometrium appears to spread beyond the endometrial/myometrium interface, making the adjacent myometrium appear artifactually bright (Figure 43-40).⁷³ A problem encountered with some (not wide-angle) transvaginal probes in gynecologic imaging is related to the limited size of the volume "box." Because of this limitation, the uterus may not be included in its entirety in a single volume. In some cases, it may be necessary to acquire 2 volumes: one for the cervix and a second for the uterine body and fundus. For this reason, the multiplanar display should be quickly reviewed immediately after acquisition to determine that portions of a structure of interest have not been left out.

Likewise, a very large adnexal mass may not be imaged completely in any single volume of data obtained transvaginally. Some of the newer transvaginal volume transducers have a very wide angle and these problems are not encountered as often. The size of the volume "box" on the abdominal volume probe is larger; a large mass may need to be imaged in this way. Similar volume size issues might also be encountered in third trimester fetuses. Another potential pitfall in 3DUS involves spatial orientation within the saved volume data. During the preliminary real-time 2D scanning, the operator must determine the left and right orientation of the structure of interest and ideally annotate that in the screen prior to volume acquisition, such as left arm, right leg, left ear, right kidney, the uterus (anteverted or retroverted), etc. If the acquired volume is manipulated prior to saving, the correct orientation of the anatomic structure of interest may not be apparent on later review of the volume. The saved volumes of structures like fetal ear, eye, extremities, or ovaries/adnexa may be annotated as well with respect to left and right, because one cannot reliably determine this from review of the volume at a later time.

Shadowing artifacts also remain a real issue, especially in fetal 3DUS. In a volume, they tend to be compounded. Most of the above-mentioned artifacts may actually create false-positive findings. A good practice is to always have more than 1 (ideally at least 3) volumes for any structure where an anomaly is suspected to avoid misdiagnosis.

It must be emphasized that 3DUS information is based on compilation and manipulation of conventional 2DUS data. Therefore, it is still prone to problems that affect 2DUS such as unfavorable maternal body habitus, motion artifacts due to fetal movement, and decreased amniotic fluid. Although 3DUS might improve our overall comprehension of anatomy, it does not make up for poor scanning technique. Poor resolution, shadowing, and poor penetration in 2DUS will most likely bring about a suboptimal image quality with 3DUS as well. ⁷¹

The most important aspect of 2D- and 3DUS training is the ability of the operator to fully master standard techniques that yield reproducible results in clinical scanning.

In our experience, the process of training medical personnel to truly understand these principles and reliably apply them in clinical practice or in a research protocol is somewhat lengthy. A comprehensive and continuous education and quality review system will have to be in place before we can expect any sonographer to reliably acquire data to be used for clinical or research purposes. The fact is that formal training in 3D/4D ultrasound is still absent in our sonography schools, residency programs, or fellowship programs. Most of the training given at the moment is from the manufacturers themselves (some of them) where ultrasound operators get primarily equipment-specific instruction that mainly revolves around instrumentation. While experience with 2D ultrasound is beneficial to understanding 3DUS, that knowledge alone is simply inadequate to perform 3DUS clinically. The time commitment to learn to appropriately perform 3DUS can be significant. In addition, the learning curve, especially when it comes to volume acquisition and analysis for quantification purposes, not just visualization (which unfortunately is where most people focus their attention), can be rather steep.

At the moment, volume ultrasound technologies are not standardized.^74,75 The process of acquisition, display, and manipulation of the volumes differs, at times significantly, among manufacturers. Even with one manufacturer, at times these processes may not be standardized across all their platforms. This reality makes the process of training even more difficult.

The above discussion clearly supports the great need for additional research and technological development. Specific acquisition methods that yield the most information for certain organ system scan plane extractions and automation, segmentation techniques (especially selection tools), organ volume nomograms, and documentation remain important areas where much research work needs to be done.

While volume ultrasound imaging techniques are more widely utilized and better understood, there are a number of emerging concepts that have the potential to significantly improve acquisition, visualization, qualitative and quantitative analysis, and documentation. Some of the most interesting ones are discussed below.

Acquisition

Here the most intriguing emerging concepts all have to do with the development of 2D matrix arrays. Their wide implementation may require a fundamental change in the way ultrasound will be practiced. The concept of standard or conventional planes will have to be expanded. Impossible views with depth will have to be included. Measurements may not be optimal if they are made blindly in one scan plane, as is common now, but will have to be correlated in a multiplanar view that truly represents the long axis (ie, long bones) or the proper measurement plane (ie, BPD, head circumference [HC]). The issue of which ultrasound technology to use when will most likely be determined by questioning how thick the ultrasound beam should be kept and which rendering algorithms should be used to optimally visualize certain anatomical planes, and not simply whether to perform a 2D or a 3D scan.

The use of volume ultrasound for interventional guidance of diagnostic and therapeutic procedures should also be expected to increase as operators become familiar with its capabilities.

There is a clear trend towards volume ultrasound miniaturization, not only transducer footprint, but of ultrasound systems as a whole. Great efforts are being made by the manufacturers to miniaturize the hardware, from custom miniaturized beam-former designs, graphic boards, mobile processing, storage, and displays to smaller footprint probes without significant trade-offs in image quality. Making volume ultrasound equipment smaller means it becomes more portable and closer to the patient's bedside and increases its clinical role and availability. Physicians or sonographers who cover large geographic areas and have to travel will not have to have a different machine in every location. A single, high-resolution, reliable, portable, laptop-based piece of equipment may adequately fulfill their needs.

Telemedicine applications of 3DUS are another new and promising area. Medically underserved, remote areas can have access to expert consultation if they can send their volumes to a referral center via Internet. Currently, there are many telemedicine setups via ISDN or even fiberoptic lines sharing mainly image and video, but the clinical quality of these acquisitions remains an issue. Given the operator skill dependency of ultrasound, another concept entertained is the remote acquisition of 2D images/video sequences and 3D volumes via a robotic arm or simply guiding the operator live via a broadband video feed precisely where and how to take the acquisition, provided they are scanning with a volume system.

Contrast agents will definitely see an increase in their role in obstetrics and gynecology, especially in gynecology applications.⁷⁶ Volume sonography has a unique role in this development. Dedicated algorithms need to be developed for specific volume applications such as tubal patency, detection of blood flow both in normal and in neovascularization, or for using tissue-specific agents such those for the lymphatic tissues, etc.

New specialized transducers for specific clinical exams are being designed. Different applications need different customized designs that take into account the specific focal point and focal range of the intended anatomical area. Examples are as follows:

1. Volume transrectal side-fire linear probes to better image the urethra and the uterus/cervix. These probes would not occupy space in the vaginal area and can be used for interventional guidance of transvaginal or transurethral procedures.

2. Very-high-frequency volume probes for first trimester embryonic ultrasound, realizing they will not have penetration.^77,78

3. Volume probes that are able to use new diagnostic and therapeutic ultrasound technologies such as elastography^{79,80, and 81} (uterus, fibroids, endometrium, pregnant cervix) or high-intensity focused ultrasound (HIFU).^{82,83,84,85, and 86} With HIFU, the feedback loop giving information on the local temperature change during focused ultrasound ablations cannot be accomplished yet by ultrasound. New tissue- and temperature-specific detection algorithms along with ablation location tracking in a volume will need to be designed.

A relatively new concept being studied is the use of 3DUS as a screening tool by improving the workflow.^{87,88, and 89} Some ultrasound laboratories are entertaining the idea of performing a significant portion of their screening scans by acquiring 3D volumes and displaying them via multiplanar or tomographic displays for offline documentation and review. The required images for the clinical protocols are all generated from the acquired volumes. Goncalves et al showed that information provided by reviewing 3D volume sets is accurate and consistent in most cases with the original 2D imaging results.⁸⁹ Another earlier large study by 2 institutions also came to a similar conclusion.⁹⁰ This is an interesting concept that will need to be validated by other investigators in different practice settings. To be implemented on a large scale, extensive training for both sonographers and especially reading physicians (who will have to generate the clinical views from the volumes) would be an important prerequisite. As equipment and training become more widely available, the use of 3DUS as a screening tool may be an important aspect of ultrasound practice, leading to shorter examination times, increased workflow efficiency, and even reduced scanning-related injuries.

Visualization

We are still visualizing 3D information with depth via flat 2D monitors. True high-definition (1080-pixel and even higher) volumetric displays capable of showing depth have been introduced in other imaging applications since 2004, but have yet to equip ultrasound machines. As demand and mass production increases in other industries we would expect to see them emerge in our systems. They would create a whole new user experience. In the meantime, stereo displays have shown real promise in better delineating anatomy acquired with volume ultrasound.⁹¹

The whole process of human-machine interaction will need to be improved. There is a need for better input devices as volume ultrasound is increasingly demanding plane and spatial manipulation. Some preliminary work has been published on kinesthetic interaction through haptic devices and is very encouraging.^{92,93, and 94} Pressure-sensitive digitized pens for tracing and force-feedback input devices for plane manipulation should soon find their way into ultrasound machines as current touch panels and hard keys make the user interface busy and cumbersome. Interactive touch-sensitive displays (gesture based and multitouch) would be expected to be play a major role in volume exploration and analysis.

Visualization of volume-rendered data is receiving additional features such as different virtual light sources to illuminate parts of the anatomy that are not necessarily perpendicular to the beam (see Figure 43-38). In addition, setups are emerging where a number of virtual cameras can display the anatomy in a fly-through and perspective rendering, which is especially useful in visualizing inside rendered hollow structures and cavities simulating endoscopic/ hysteroscopic visualization.^95,96

Qualitative and Quantitative Post-processing

Automation is the new frontier. Algorithms for automatic border selection and volume quantification,^22,97 scan plane extraction automation, and quantification of the spatial relationships between views of the heart,^50,74 or even built-in automation for computerized quality control testing,⁹⁸ are emerging fields where a lot of effort is being directed at the moment. Image fusion with CT and MR imaging is still in the experimentation phase, but should see a boost now that standard planes can be generated with 3DUS.

Selection tools will need to be drastically improved with classic tools used in the graphics industry, such as preselected shapes, Bezier curves, editable paths, etc.

Pattern recognition, shape analysis, and quantification have seen some very promising preliminary results and should provide unique insight on normal and abnormal organ development.^66,69 Perfusion quantification remains problematic. Weather using B-flow or true power or color Doppler quantification,⁹⁹ with or without contrast agents,⁷⁶ perfusion studies would have a significant impact on our understanding of normal and abnormal fetal-placental physiology.

Documentation

Image and volume documentation remains an important issue. Increasingly, manufacturers are offering the file extension .raw as an option (or even by default) to save the collected 2D or 3D/4D-generated multiplanar or rendered images. The raw file format contains minimally processed data from the imaging equipment. Raw image files are sometimes called digital negatives, as they fulfill the same role as film negatives in traditional chemical photography, that is, the negative is not directly usable as an image, but has all of the information needed to create an image. The implication for us is that, while we are supplied with higher-resolution images, a raw file at the moment still can only be opened with that manufacturer's proprietary software as it is encoded as such. This creates many problems for long-term storage and sharing of this information. This problem is the same for many other imaging industries. For example, one standard raw format has been proposed (with a free converter to it) issued by Adobe Systems called the the "Digital Negative" (DNG) format. This is a free and publicly available archival format for raw files. Of course, the manufacturers will have to adopt this format as well.

Volume ultrasound technologies are being developed by different manufacturers, and their file format and specifications remain quite different and incompatible with each other. Making the 3DUS file formats proprietary makes it obligatory to open and manipulate them only in a computer with software supplied from that particular manufacturer. The DICOM Working Group 12-3D (WG 12-3D) was established to define a complete image and storage specification for 3D and 4D data supporting open exchange between different vendors as well as multimodality (ultrasound, CT, MR imaging). The goal is to prepare an enhanced format for 3DUS.¹⁰⁰ This would be a huge step forward for the ultrasound community and will greatly facilitate documentation, storage, and networking, as well as training and research.

Highlighted References