INSTRUMENTATION DESIGN FOR ULTRASONIC IMAGING

Transcription

1 Source: STANDARD HANDBOOK OF BIOMEDICAL ENGINEERING AND DESIGN CHAPTER 25 INSTRUMENTATION DESIGN FOR ULTRASONIC IMAGING Kai E. Thomenius GE Corporate Research and Development, Schenectady, New York 25.1 INTRODUCTION BEAM FORMATION BASIC CONCEPTS SIGNAL PROCESSING AND SCAN 25.3 TYPICAL SYSTEM BLOCK CONVERSION DIAGRAM SUMMARY INTRODUCTION The purpose of this chapter is to show how piezoelectric transduction, sound wave propagation, and interaction with scattering targets are taken advantage of in image formation with an ultrasound instrument. These instruments have evolved over the last 40 years from relatively simple handmoved scanners built around an off-the-shelf oscilloscope to rather sophisticated imaging computers. Much technology has been perfected during this evolution. For example, transducers have grown from circular single-element probes to precision arrays with more than 1000 elements. With better frontend electronics, the operating frequencies have increased as weaker echoes can be handled. As the gate counts of VLSI ASICs* have increased, the numbers of processing channels in array-based systems have risen. With the introduction of reasonably low cost high-speed (20 to 40 MHz) 8- to 12- bit analog-to-digital converters, digital beam formation has become the standard. The organization of this chapter is based on the discussion of a block diagram of a generalized ultrasound system. Each component of the block diagram will be reviewed in considerable detail. Different design approaches for the various blocks will be reviewed and their advantages and disadvantages discussed. Finally those areas of the block diagram that are targets of significant current research are summarized BASIC CONCEPTS Image Formation Image formation in medical ultrasound is accomplished by a pulse-echo mechanism in which a thin ultrasound beam is transmitted and the echoes generated by the interaction of that beam with * Very large-scale integration application-specific integrated circuits. 25.1

2 25.2 DESIGN OF MEDICAL DEVICES AND DIAGNOSTIC INSTRUMENTATION scattering targets are received by a transducer or a set of transducer elements. The transmit and receive processing used to create this beam is referred to as beam formation. Because of its central role in ultrasonic imaging, beam formation will be discussed in detail later on. The strength of the received echoes is usually displayed as increased brightness on the screen (hence the name for the basic ultrasonic imaging mode, B-mode, with B for brightness). A twodimensional data set is acquired as the transmitted beam is steered or its point of origin is moved to different locations on the transducer face. The data set that is acquired in this manner will have some set of orientations of the acoustic rays. The process of interpolating this data set to form a TV raster image is usually referred to as scan conversion. With Doppler signal processing, mean Doppler shifts at each position in the image can be determined from as few as 4 to 12 repeated transmissions. The magnitudes of these mean frequencies can be displayed in color superimposed on the B-mode image and can be used to show areas with significant blood flow Physical Constants and Typical System Operating Parameters It may be useful to consider typical system operating parameters. The following table lists some physical constants that help define the operating parameters of today s systems: Typical attenuation in tissue Speed of sound in tissue 0.5 db/cm MHz for one-way travel 1540 m/s (or approximately 13 µs/cm for roundtrip travel) One of the challenges of ultrasonic imaging relates to that very high attenuation. To put this in numerical terms, a typical 5-MHz transducer is expected to penetrate about 10 cm. Thus, the signal leaving the transmitting transducer will undergo attenuation in the order of 25 db before it reaches a scattering site. At that point, a small fraction of the energy will be reflected; let us say the reflected echo will be another 30 db down, and the return will bring about another 25 db. Thus the entire attenuation has been about 80 db. Needless to say, there is a strong need for careful low-noise designs for ultrasound front ends. The following table gives some of the typical system design parameters commonly used for B- mode imaging: Transducer frequencies 2 15 MHz Transducer bandwidth 50 90% fractional bandwidth Typical depths of penetration 18 cm (abdominal imaging) 16 cm (cardiac imaging) 4 cm (small parts and peripheral vascular imaging) Time to acquire one 20-cm acoustic ray ~ 260 µs Pulse repetition frequency (PRF) 4 khz Typical number of rays in an image Data acquisition frame rates frames/s Because of frequency-dependent attenuation, applications with greater penetration requirements use the lower transducer frequencies. The instrument parameters have been selected or have evolved to their current values as the manufacturers have optimized their instruments to the needs of the various clinical applications. Given the compromises that have to be made, resolution of scanners is been limited, ranging from roughly 0.5 mm with 7- to 10-MHz transducers to about 2 to 4 mm with 2.5- to 3.5-MHz transducers. The penetration at which these resolutions can be achieved is about 3 cm for the higher frequencies and 15 or more cm for the lower frequencies. Whether or not this performance can be achieved on any given patient is dependent on factors such as uniformity of speed of sound, which is highly patient dependent (O Donnell, 1988). The degree of and correction for sound speed

3 INSTRUMENTATION DESIGN FOR ULTRASONIC IMAGING 25.3 variations in ultrasound systems continues to receive much attention (for example, Fink, 1992; Flax, 1988; Li, 1995; Nock, 1988; O Donnell, 1988; Trahey, 1988; Zhu, 1993a and 1993b; Krishnan et al., 1996; Rigby, 2000; Silverstein, 2001) Clinical Applications B-mode imaging has found numerous uses in today s clinic (Goldberg, 1990; Sarti, 1987). Some of these are: Abdominal imaging Cardiology Obstetrics Peripheral vasculature Identification of tumors, cysts in liver, kidneys Valvular insufficiency (flail leaflet), myocardial dyskinesis, septal defects, congenital malfor mations Fetal growth, congenital malformations Extent of plaque, blood vessel tortuosity Many of these diagnoses are based on relatively gross anatomical information that is available from the ultrasound image. In addition there is a large amount of additional information imparted to the echoes by the scattering process. Some of this information is displayed on B-mode images and is of major value to the clinician. This information is critical to diagnoses such as diffuse disease processes, the identification of tumors, quality of myocardial dynamics, and so forth. It is for these reasons the signal integrity and retention of maximum dynamic range is of key value in the image formation process. Sales of ultrasound instruments are divided among the four areas listed above roughly as follows (source: Klein Biomedical Consultants): Radiology 39% Cardiology 35% Obstetrics/gynecology 16% Peripheral vasculature 5% This gives a rough idea of the level of utilization in the several areas; however, it also should be noted that the marketplace is currently undergoing significant changes with the reform activity in health care. Also, there is much overlap between the segments. For example, many radiologists do perform obstetric or peripheral vascular examinations and echocardiologists perform increasing amounts of peripheral vascular work. In terms of instrumentation sales, these are believed to be approximately $2.5 billion in year Classifications of Ultrasound Instruments Ultrasonic instruments can be classified in many of different ways (Christensen, 1988). Among these are: Types of electronic beam steering: Phased arrays versus steering by element selection Clinical applications: See Sec Nature of beam formation: Analog versus digital Portability Console-based systems versus hand-helds

4 25.4 DESIGN OF MEDICAL DEVICES AND DIAGNOSTIC INSTRUMENTATION FIGURE 25.1 Range of ultrasound scanners. As can be seen the ultrasound scanners come in a range of physical sizes. (Courtesy of GE Medical Systems.) With respect to steering methods, the great majority of instruments (perhaps more than 95 percent) sold today are electronically (as opposed to mechanically) steered. Phased-array systems are dominant in echocardiographic applications where aperture size is limited by rib spacings, while the other beam steering methods are more often used in radiologic and obstetric and gynecological examinations. The latter grouping is sometimes referred to as general imaging ultrasound. The shift to digital beam formation is accelerating, and most of the instruments sold have digital beam formers. This is true even for the lower-price points. Figure 25.1 shows a range of ultrasound scanners TYPICAL SYSTEM BLOCK DIAGRAM The block diagram in Fig shows the signal processing steps required for B-mode image formation. The actual implementations vary considerably among manufacturers and the types of systems. For example, the grouping of functions might be different from one system to the next, depending on the choices made by the system designers; however, the basic functionality shown by each block has to be there. One point that the block diagram may not convey adequately is the degree of duplication of functions in today s systems For example, in systems with 128 processing channels, there will usually be 128 pulsers, 128 transmit/receive switches (T/R switches), and so forth. In such systems the use of large-scale integration and application specific integrated circuits (ASIC s) is highly important for cost and space reduction. For the most part, the block diagrams for digital and analog beam formers are quite similar, although there will usually be a large number of additional support circuitry required for synchronization, interpolation, and decimation of the sampled waveforms, and so forth. Depending on the particular system implementation, the full RF bandwidth will be retained through the top part of the block diagram on a typical analog design. The class of heterodyned systems performs frequency mixing at the beam former level, and the signal spectrum is shifted to either an intermediate frequency or all the way to the baseband. With digital beam formers, A/D conversion occurs after the variable gain stages. The digital beam former systems can be designed with similar heterodyning approaches, although there are many different approaches to delay generation. In addition to the digital and analog forms of beam formers, it is also possible to use charge-coupled devices for this purpose and these, in fact, are receiving considerable attention at the present time.

5 INSTRUMENTATION DESIGN FOR ULTRASONIC IMAGING 25.5 FIGURE 25.2 Block diagram of a typical ultrasound system. The blank circles represent points at which user control is introduced. The following paragraphs introduce and describe briefly the functions of the most important blocks in Fig The most important of these will receive greater attention in the remainder of the chapter B-Mode Transducers The mode of transduction in ultrasound systems takes advantage of the piezoelectric characteristic of certain ceramics. There are several types of transducers currently in use, and the nature of processing of acoustic data is different among them. These differences are highlighted in later sections as appropriate. The piezoceramics can be fabricated in a number of ways to perform B-mode imaging. In its simplest form, the B-mode transducer is a circular single-element transducer with a fixed geometric focus. To a large degree this type of a transducer has been replaced today by more sophisticated multielement transducers; however, there are still commercial systems where single-element transducers are used. The next more complicated designs are annular arrays, which are also circular

6 25.6 DESIGN OF MEDICAL DEVICES AND DIAGNOSTIC INSTRUMENTATION but, as the name implies, are composed of several (4 to 12) rings. These are the transducers most often used with mechanically steered systems (to be discussed later). Both the single-element and annular array transducers usually have a concave curved transmitting surface or an acoustic lens to focus the acoustic energy at a given location. The next broad transducer category is that of linear arrays, which are built by dividing a piezoceramic strip into a large number of line-source-like elements. The number of elements in such arrays can exceed 200, although 128 is a typical number. A variation of the linear arrays is curvilinear arrays which are built around a curved surface. With linear arrays, the acquired image is rectangular in shape, while with curvilinear arrays (and single element, annular arrays, and certain linear arrays), it is sector shaped. Both linear and curvilinear arrays either have a lens or are curved to focus the ultrasound in the plane perpendicular to the imaging plane. With both of these types of arrays, focusing in the image plane is done electronically. Finally, in order to improve on slice thickness, some form of elevation focusing is being introduced into today s systems. This is realized by the use of multiple rows of elements which form a twodimensional (2-D) array of elements. Arrays that are connected to form a symmetrical aperture about the center line are sometimes referred to either as 1.25-D or 1.5-D arrays depending on whether improved slice thickness in the elevation direction is achieved by increasing the aperture size or by performing electronic focusing with those rows of elements. The transducer or transduction itself has received considerable attention from researchers because of its critical position in the signal processing sequence (Hunt, 1983). Much progress has been made in the areas of sensitivity, bandwidth, and the use of composite materials. Bandwidths of today s transducers can exceed 80 percent of the center frequency. This gives the system designer additional flexibility with the available imaging frequencies and allows the optimization of image quality among the various imaging modes as well as over a larger cross section of patients. There is some potential of increasing the transmitted signal bandwidths to above 100 percent with a new class of transducers referred to as single-crystal relaxors, about to be introduced into the marketplace Pulser, T/R Switch, Variable Gain Stage (or TGC Amplification) This group of blocks is among the most critical from the analog signal-to-noise ratio point of view (Schafer and Lewin, 1984; Wells, 1977). The piezoceramic array elements are energized with the pulser, and the transduction occurs as has been described. With the newer generation of instruments, the excitation waveform is typically a short square-wave burst with one to three cycles. Earlier generations (pre-doppler) of instruments tended to use shock excitation with very wide bandwidths. The transducer element, with its limited bandwidth, filters this signal during both transmission and reception to a typical burst. The pulser voltages used vary considerably but values around 150 volts are common. Use of a short burst (say with a bandwidth in the 30 to 50 percent range) can give the system designer the ability to move the frequency centroid within the limits of the transducer bandwidth. In some imaging modes such as B mode, the spatial-peak temporal-average intensity (I spta, an FDA-regulated acoustic power output parameter) value tends to be low; however, the peak pressures tend to be high. This situation has suggested the use of coded excitation, or transmission of longer codes that can be detected without loss of axial resolution. In this manner the average acoustic power output can be increased and greater penetration depth be realized. The T/R switches are used to isolate the high voltages associated with pulsing from the very sensitive amplification stage(s) associated with the variable gain stage [or time gain compensation (TGC) amplifier]. Given the bandwidths available from today s transducers (80 percent and more in some cases), the noise floor assuming a 50-Ω source impedance is in the area of few microvolts rms. With narrower bandwidths this can be lowered but some imaging performance will be lost. If the T/ R switch can handle signals in the order of 1 volt, the dynamic range in the neighborhood of 100+ db may be achieved. It is a significant implementation challenge to have the noise floor reach the thermal noise levels associated with a source impedance; in practice there are a good number of interfering sources that compromise this. In addition, some processing steps such as the T/R switching cause additional losses in the SNR.

7 INSTRUMENTATION DESIGN FOR ULTRASONIC IMAGING 25.7 The TGC stages supply the gain required to compensate for the attenuation brought about by the propagation of sound in tissue. During the echo reception time, which ranges from 40 to 240 µs, the gain of these amplifiers is swept over a range approaching 60 to 70 db, depending on the clinical examination. The value of this gain at any depth is under user control with a set of slide pots often referred to as the TGC slide pots. The dynamic range available from typical TGC amplifiers is in the order of 60 db. One can think of the TGC amplifiers as providing a dynamic range window into the total range available at the transducer. This is illustrated in Fig FIGURE 25.3 Front-end block diagram with signal processing steps and corresponding dynamic ranges. It is interesting to note that the commercial impact of multichannel ultrasound instruments is such that a special purpose TGC amplifier IC has been developed for this function by a major integrated circuit manufacturer (Analog Devices in 1990) BEAM FORMATION Beam formation can be considered to be composed of two separate processes: beam steering and focusing (Macovski, 1983). The implementation of these two functions may or may not be separated, depending on the system design. Focusing will be discussed first Focusing Analogously to optics, the spatial variation in system sensitivity can be modified by the action of focusing on the transmitted acoustic beam and, during reception, on its echoes. One can view focusing as the modification of the localized phases (or, more correctly for wideband systems, time shifts) of the acoustic beam so as to cause constructive interference at desired locations. One simple way to accomplish focusing is by curving the transducer element so as to form a phase front that, after traveling a defined distance, will cause the beam to add constructively at a desired focal point. With transducer arrays, the formation of the desired phase front during transmission can be accomplished by electronically altering the sequence of excitation of the individual elements. Similarly, during reception, the signals from the array elements can be routed through delay lines of appropriate lengths so that echoes from specific locations will have constructive interference (Thurstone, 1973). These processes are shown schematically in Fig As suggested by Fig. 25.4, the echoes from a point source will have a spherical wavefront. The center elements of the array will receive these echoes at first while the outer elements will receive them last. To compensate for this and to achieve constructive interference at the summer, the center elements will be given the longest delays, as suggested by the length dimension of the delay lines. The calculation to determine the differential delays among the received echoes is straightforward. An attractive formalism for expressing the array-based focusing in mathematical terms is due to Trahey (1988). The total transmitted pressure wave T(t) at a point p can be expressed as a sum of the contributions from N array elements as follows:

8 25.8 DESIGN OF MEDICAL DEVICES AND DIAGNOSTIC INSTRUMENTATION FIGURE 25.4 Schematic of focusing during reception. The echoes from a point source (at 40 mm) are shown impinging on a transducer array. The difference in the reception times is corrected by the delay lines. As an example, the echo will be received first by the center elements. Hence, their delays are the longest. (25.1) where A T (n) = pressure amplitude contribution of the nth element of the array at point p r(n) = distance from the nth element to the focus S(t) = waveshape of the pressure pulse generated by any given element of the array t T (n) = focusing time delay for element n shown as the length of delay lines in Fig c = speed of sound in the medium Assuming that at location p there is a point target with reflectivity W p, then the signal after the summer in Fig can be described by (25.2) where A T (n) = pressure amplitude contribution of the nth element to echoes from point p t R (t) = receive focusing delay for element n and T(t) is given by Eq. (25.1). The remaining parameters of Eq. (25.2) were defined in Eq. (25.1). It should be noted that the A T (n) and A R (n) terms in Eq. (25.1) and (25.2) will, in general, be different since the transmit and receive operation need not be symmetric. The terms include tissue attenuation, element sensitivity variation, and transmit or receive apodizations. It might be useful at this point to discuss several methods by which the receive delays for either focusing or beam steering are implemented. The previous paragraph refers to the use of delay lines for this purpose. Analog delay lines are an older method, albeit a very cost-effective one. However, lumped-constant delay lines do suffer from several limitations. Among these is the limited bandwidth associated with longer delay lines. Delays needed for focusing for most apertures are less than 0.5 µs;

9 INSTRUMENTATION DESIGN FOR ULTRASONIC IMAGING 25.9 however, for phased-array beam steering (see below) they may be as long as 8 µs for largers apertures required for 2.5- to 3.5-MHz operation or up to 5 µs required for 5- to 7-MHz operation. Delay lines suitable for the latter case are relatively expensive. In addition, there are concerns about the amplitude variations with tapped delay lines as different taps are selected, delay uniformity over a production lot, and delay variations with temperature. In response to these difficulties, there has been a major migration to digital beam formation over the last 10 years (Thomenius, 1996). An alternative method of introducing focusing delays for both analog and digital beam formers is by heterodyning (for example, Maslak, 1979). This is usually done in conjunction with mixing the received signal with a lower local oscillator frequency with the goal of moving the received energy to a different location on the frequency spectrum. If the phase of the local oscillator is varied appropriately for each of the array signals, the location of constructive interference can be placed at a desired location. The limitations of this are associated with the reduced bandwidth over which the delay correction will be accurate and the reduced range of phase correction that is possible. Finally, as noted above, focusing (and beam steering) can be accomplished by relatively straightforward digital techniques in a digital beam former. A number of different methods of digital beam former implementation have been published in the sonar and ultrasound literature (Mucci, 1984; Steinberg, 1992). Figure 25.5 shows the formation of the focal region for a 20-mm-aperture circular transducer with a geometric focal distance of 100 mm and excited with a CW signal of 3.0 MHz. At the lefthand side of the pressure profile, the rectangular section from -10 to 10 mm corresponds to the pressure at the transducer aperture. In the near field, there are numerous peaks and valleys corresponding to FIGURE 25.5 Spatial cross-sectional pressure distribution due to a circular transducer with a diameter of 20 mm, frequency of 3.0 MHz, and a radius of curvature of 100 mm. The left-hand side corresponds to the transducer aperture. All spatial dimensions are in millimeters; the x and y axes are not to scale. This pressure distribution was determined by the use of angular spectrum techniques. (Schafer, 1989.)

10 25.10 DESIGN OF MEDICAL DEVICES AND DIAGNOSTIC INSTRUMENTATION FIGURE 25.6 The 6- and 20-dB beam contours for the beam generated by a 3.0- MHz, 19-mm-aperture, 100-mm-focus transducer undergoing CW excitation. The x and y axes are to scale so that one can get a sense of the beam dimensions with respect to the depth of penetration. areas where there is partial constructive and destructive interference. As one looks closer to the focal region, these peaks and valleys grow in size as the areas of constructive and destructive interference become larger. Finally, at the focal point the entire aperture contributes to the formation of the main beam. One way of assessing the quality of a beam is to look at its beamwidth along the transducer axis. The 6- and 20-dB beamwidths are plotted on Fig It is important to recognize that the beamwidths shown are those for a circular aperture. Because of the axial symmetry, the beamwidths shown will be achieved in all the planes, i.e., in the imaging plane as well as plane perpendicular to it (this plane is often referred to as the elevation plane, from radar literature). This will not be the case with rectangular transducers. With rectangular transducers, the focusing in the image plane is done electronically, i.e., in a manner similar to that for annular arrays. However, in the elevation plane, the focusing in today s systems is done either by a lens or by the curvature of the elements. In such cases the focal location will be fixed and cannot be changed electronically. Remedying this limitation of rectangular transducers is currently an active area of study. The introduction of the so-called elevation focusing will be discussed in greater detail in a later chapter. There is considerable diagnostic importance that has to be attached to the 20-dB and higher beamwidths. Sometimes the performance at such levels is discussed as the contrast resolution of a system. The wider the beam is, say at 40 db below the peak value at a given depth, the more unwanted echoes will be brought in by these sidelobes. Small cysts and blood vessels may be completely filled in by such echoes. Also, if a pathological condition alters the backscatter strength of a small region by a modest amount, this variation may become imperceptible because of the acoustic energy introduced by the sidelobes. With array-type transducers, the timing of the excitation or the delays during reception can be varied, thereby causing a change in the focal location. This is demonstrated in Fig. 25.7, where three different 6-dB profiles are shown. During transmission, the user can select the focal location as dictated by the clinical study being performed. There are operating modes sometimes referred to as composite imaging modes in which the final image is a composite of the data acquired during transmission and reception from several distributed focal locations. Not only can one change the transmit focal location but also the aperture size and transmit frequency between the transmissions. With this approach, one can achieve superior image uniformity at the expense of frame rate, which will be decreased by the number of transmissions along a single look angle.

11 INSTRUMENTATION DESIGN FOR ULTRASONIC IMAGING FIGURE 25.7 Three 6-dB beam profiles for an array-type transducer. The radii of curvature for the three cases are 30, 50, and 80 mm. During reception there is yet another possibility to improve the performance of the system. As the transmitted wavefront travels away from the transducer, the delays introduced by the focusing can be varied, thereby changing the receive focus with the location of the wavefront. This approach is referred to as dynamic focusing and is now a standard feature of most systems. With analog delay lines, dynamic focusing is often implemented with tapped delay lines by changing the tap selection as the required focal delays change. Implementation of dynamic focusing by this method does add a number of challenges to the system designer, among which are the introduction of switching noise in the case of analog delay lines Beam Steering There are a number of different methods of beam steering currently in use. These can be grouped into three categories: 1. Mechanical 2. Element selection 3. Phased array The major implication of the selection of the beam steering approach is in the cost of the instrument. Mechanically steered systems tend to be the simplest and hence the least expensive while the phased arrays are the most expensive. Some imaging systems incorporate all three types; the great majority of recent vintage scanners have the latter two types of beam steering. The following paragraphs will discuss the relevant features of each of the three types. Mechanical Steering. The simplest method of beam steering is to use a mechanism to reorient a transducer (usually a circular aperture) to a predetermined set of orientations so as to capture the required two-dimensional data set. This approach was dominant at first; however, in the last 15 years, electronically steered systems have become, by far, the most popular. Mechanical systems usually use either a single-element transducer or an annular array transducer (Fig. 25.8). The former will have a fixed focus while the latter does allow the focal point to be moved electronically. This will be discussed more fully later.

12 25.12 DESIGN OF MEDICAL DEVICES AND DIAGNOSTIC INSTRUMENTATION FIGURE 25.8 Sketch of a mechanically steered transducer. The transducer is isolated by a nosepiece and bathes in a coupling fluid. Included in the sketch are the ray paths due to the radius of curvature. A motor or a motor drive (not shown) moves the transducer back and forth. Because of that motion, this design is sometimes referred to as a wobbler. There are a number of very attractive aspects to mechanical systems. Among these are low cost and the ability to focus the sound beam electronically in all planes. The low cost arises from the relatively low cost associated with the mechanisms used to move the transducer in comparison to the multielement transducer arrays and supporting electronics needed with electronic beam steering. The ability to focus the acoustic energy in all planes is a unique advantage, since most mechanically steered systems use either single-element or annular array type transducers. With the annular arrays, one has the capability to move the focus electronically in all planes, as opposed to the electronically steered arrays, which are usually rectangular in shape and will have electronic focusing in only one plane. The number of transducer elements in an annular array is usually less than 12, typically 6 or 8. With electronically steered arrays, the element count can go as high as 192 or more. As a consequence, the costs tend to be higher. With mechanical steering, one can vary the acoustic line density by changing the speed of rotation of the transducer. The system designer has the choice of maintaining a constant pulse repetition frequency but winding up with a variable line density or, by making the pulse repetition timing a function of the location of the transducer at any point of the sweep and thereby achieving uniform line density. The best alternative may be to try to achieve a linear sweep with the transducer across the area to be imaged at the same time that the transducer firing is kept as a function of the transducer angle. The last choice makes the challenges of scan conversion more tolerable. Some of the drawbacks associated with mechanical steering involve the inertia associated with the transducer, the mechanism, and the fluid within the nosepiece of the transducer. The inertia introduces limitations to the frame rate and clearly does not permit random access to look angles as needed (the electronically steered approaches supply this capability). The ability to steer the beam at will is important in several situations but most importantly in Doppler applications. Further, electronic beam formation affords numerous advanced features to be implemented such as the acquisition of multiple lines simultaneously and elimination of the effects due to variations in speed of sound in tissue. Steering by Element Selection. Another relatively low cost beam-steering approach involves steering of the beam by element selection. In this approach one doesn t strictly steer the beam but rather changes the location of its origin, thereby achieving coverage over a two-dimensional tomographic

13 INSTRUMENTATION DESIGN FOR ULTRASONIC IMAGING FIGURE 25.9 Steering by element selection for a curvilinear array. The beam will shift to a new location as the center of the active aperture is shifted over the entire array. slice. This method is applied with both linear and curvilinear arrays. Figure 25.9 shows the application in the case of curvilinear arrays. For this particular case, the two-dimensional image will be sector shaped; with linear arrays, it will, of course, be rectangular. This is a relatively low cost approach since aside from the multiplexing required for element selection, the electronics required to accomplish beam formation are merely the focusing circuitry. The line densities achievable with this mode of beam steering are not as variable as with mechanical steering, since it they will be dependent on element center-to-center spacing. There are methods by which one can increase the achieved line density. Figure 25.9 shows an acquisition approach sometimes referred to as full stepping. The line density with full stepping will be equal to the element density since the beam center will always be at the junction between two elements. It is possible to change the sizes of the transmit and receive apertures and thereby change the transmit and receive beam centers. This changes the effective location of the resultant beam and introduces the possibility of an increased line density. Half- and even quarter-stepping schemes exist, although care has to be taken that the resulting beam travels along the expected path. Steering with Phased Arrays. The most complicated form of beam steering involves the use of phased-array concepts derived from radar (for example, Steinberg, 1976; Thurstone, 1973; Thomenius, 1996). Most ultrasonic phased-array transducers have between 64 and 128 elements. Transmit beam steering in phased-array systems is achieved by adding an incremental delay to the firing time of each of the array elements that is linearly related to the position of that element in the array. Similarly, during reception the delay that is applied to each of the echoes received by the array elements is incremented or decremented by a position-dependent factor. This differential time delay t is given by (25.3) where x n is the location of the array element n and is the desired beam steering angle. The application of such a delay increment during reception is illustrated in Figure Since the beam

14 25.14 DESIGN OF MEDICAL DEVICES AND DIAGNOSTIC INSTRUMENTATION FIGURE Beam steering in a phased-array system during reception. A linearly increasing time delay differential is introduced for each of the delay lines to correct for the linear time difference in the arrival times. steering angle is such that the echoes will reach the array elements toward the bottom of the figure first, the longest delays will be imposed on the echoes from those elements. Since the wavefront is linear, the arrival times of the remaining echoes have a linear relationship, hence the linear decrement on the delays from one element to the next. In addition to beam steering, one can combine focusing and beam steering delays in the same process. This is illustrated in Figure Echoes from a near-field point source (at 40 mm) are shown arriving at the array elements. The arrival times of the echoes have a nonlinear component so FIGURE Schematic of beam steering and focusing during reception in a phased array type system. In addition to the focusing correction (also shown in Fig. 25.4), phased-array systems add a linearly increasing time shift to the receive delay lines to achieve constructive interference in the desired direction.

15 INSTRUMENTATION DESIGN FOR ULTRASONIC IMAGING the delay lines cannot compensate with a simple linear increment as in Figure It can be shown easily that the differential time delay between channels can be determined from the law of cosines. As the process of steering and focusing is repeated for a sequence of look angles, a sector-shaped image data set is acquired. The line density available from phased-array scanners is not as restricted as with curvilinear arrays, but some limitations exist in certain systems, due to the relatively large size of delay increments available in tapped delay lines. Heterodyned systems and digital beamformers have less limitations in this area. All linear and curvilinear array systems have limitations or design constraints associated with the existence of grating lobes, which are due to leakage of acoustic energy in unwanted angles. It turns out that for certain larger center-to-center array element spacings, there will be constructive interference at look angles other than the main beam. This difficulty is particularly serious for the case of phased-array systems because of the need for beam steering. It turns out that the grating lobes move with the steering angle and can be brought into the visible region by the simple act of beam steering. Grating lobes can be completely avoided by keeping the center-to-center spacing at one-half the wavelength at the highest contemplated operating frequency. [It turns out this is completely analogous to the familiar sampling theorem, which states the temporal sampling has to occur at a frequency that is twice that of the highest spectral component of the signal being processed (Steinberg, 1976).] This has the drawback of forcing the use of a larger number of array elements and their processing channels. This, and the expensive processing required for each channel, causes the phased-array systems to be more expensive than the other types Harmonic Imaging A recent development in the area of B-mode imaging is that of imaging of the harmonics generated during propagation of acoustic waves in tissue (Averkiou, 1997; Jiang, 1998). While all the discussion so far has assumed that the propagation of these waves is linear, this is actually not the case. There is a difference in the speed of sound in the compressional and rarefactional parts of the acoustic pressure wave. As a consequence, the positive half of a propagating sine wave will move faster than the negative half; this results in the formation of harmonic energy. An image formed from such harmonics will be superior to that from the fundamental part of the spectrum because of reduced reverberant energy and narrower main beam. The acceptance of this form of imaging has been so rapid that in certain clinical applications (e.g., echocardiology) harmonic imaging is the default operating mode. From the point of view of beam former design, there is relatively little that needs to be done differently other than developing the ability to transmit at a lower frequency while receiving at twice the transmit frequency Compression, Detection, and Signal Processing Steps The sequence of the processing steps between the beam former and the scan conversion is different among the various commercial systems but the goals of the steps remain the same. The beam former output will be a wideband RF, an IF, or a complex baseband signal, which will usually be bandpass filtered to reduce out-of-band noise contributions. In systems with very wideband processing, frequency diversity techniques (e.g., split spectrum processing) can be brought into play to try to reduce the impact of coherent interference or speckle. With most of today s systems, there is a logarithmic compression of the amplified signal after beam formation amplification. The goal of this is to emphasize the subtle gray level differences between the scatterers from the various types of tissues and from diffuse disease conditions. There are a number of ways that envelope detection has been implemented. In purely analog approaches, simple full-wave rectification followed by a low-pass filtering has been shown to work quite well. It is also possible to digitize the RF signals earlier in the processing chain, perform the

16 25.16 DESIGN OF MEDICAL DEVICES AND DIAGNOSTIC INSTRUMENTATION compression and detection processes digitally, and use quadrature detection to determine the signal envelope SIGNAL PROCESSING AND SCAN CONVERSION As has been noted earlier, the image data are acquired on a polar coordinate grid for sector scanners (e.g., mechanically steered, curvilinear arrays, and phased-array systems) and in a rectangular grid for linear arrays. It is clearly necessary to convert this image data into one of the standard TV raster formats for easier viewing, recording, computer capture, etc. This is performed in a module in most systems referred to as the scan converter. The major function of the scan converter is that of interpolation from, say, a polar data grid to that of the video pixel space. Given the data rates required, it is a challenging task but one that most commercial systems appear to have well under control. The early scan converters used several clever schemes to accomplish the required function. For example, with sector scanners, certain systems would modulate the A/D converter sampling rate by the look angle of the beam former in such a way that every sample would fall onto a raster line. Once the acoustic frame was completed, and the data along each horizontal raster line was read out, simple one-dimensional interpolation was performed (Park, 1984). The need for more sophisticated processing brought about two-dimensional interpolation methods. Among the first in this area were Larsen et al. (Larsen, 1980), whose approach involved the use of bilinear interpolation. With the Larsen et al. approach, the sampling rate along each scan line was held at a uniform value which was high enough meet the sampling theorem requirements. With two axial samples along two adjacent acoustic lines, the echo strength values at each of the pixels enclosed by the acoustic samples were determined. This could be done by either (1) interpolating angularly new axial sample values along a synthetic acoustic ray that traversed through the pixel in question and then performing an axial interpolation along the synthetic ray or (2) interpolating a new axial sample along both real acoustic rays and then performing the angular interpolation. This basic approach has become the most widely used scan conversion method among the various manufacturers and seems to have stood the test of time. More recently, researchers have continued to study the scan conversion problem with different approaches. For example, Berkhoff et al. have evaluated fast algorithms for scan conversion, i.e. algorithms which might be executed by software as opposed to dedicated hardware. Given the rapid trend to faster, more powerful, and cost-effective computers and their integration with ultrasound systems, it is likely that more of the scan conversion function will be done in software. Berkhoff et al. recommend two new algorithms which they compare with several conventional interpolators (Berkhoff, 1994). With the speed of computers improving at a steady pace, these approaches are increasingly attractive (Chiang, 1997). In other work, with the cost of some of the hardware components such as A/D converters coming down, oversampling the acoustic line data may permit the replacement of bilinear interpolation with simple linear interpolation (Richard and Arthur, 1994). Oversampling by a factor of two along with linear interpolation was found to be superior to bilinear interpolation under certain specific circumstances. It is clear there is additional work in this area yet SUMMARY This chapter has reviewed the fundamentals of the design of ultrasound scanners with a particular focus on the beam formation process. Different types of image data acquisition methods are described.

17 INSTRUMENTATION DESIGN FOR ULTRASONIC IMAGING REFERENCES Averkiou, M. A., Roundhill, D. N., and Powers, J. E., (1997), A new imaging technique based on the nonlinear properties of tissues, IEEE Ultrasonics Symposium Proceedings, pp Berkhoff, A. P., Huisman, H. J., Thijssen, J. M., Jacobs, E. M. G. P., and Roman, R. J. F. (1994), Fast scan conversion algorithms for displaying ultrasound sector images, Ultrasonic Imaging, 16: Chiang, A. M., and Broadstone, S. R. (1997), Portable Ultrasound Imaging System, U.S. Patent 5,590,658, issued January 7. Christensen, D. A. (1988), Ultrasonic Bioinstrumentation, Chap. 6, John Wiley & Sons, New York. Fink, M. (1992), Time reversal of ultrasonic fields: Part I Basic Principles, IEEE Trans. Ultrason. Ferroelec. Freq. Contr., 39: Flax, S. W., and O Donnell, M. (1988), Phase aberration correction using signals from point reflectors and diffuse scatterers: Basic principles, IEEE Trans. Ultrason. Ferroelec. Freq. Contr., 35: Goldberg, B. B., and Kurtz, A. B. (1990), Atlas of Ultrasound Measurements, Year Book Medical Publishers, Chicago. Hunt, J. W., Arditi, M., and Foster, F. S. (1983), Ultrasound transducers for pulse-echo medical imaging, IEEE Trans. Biomed. Eng., 30: Jiang, P., Mao, Z., and Lazenby, J. (1998), A New Tissue Harmonic Imaging Scheme with Better Fundamental Frequency Cancellation and Higher Signal-To-Noise Ratio, Proceedings, IEEE Ultrasonics Symposium. Krishnan, S., Li, P.-C., O Donnell, M. (1996), Adaptive compensation of phase and magnitude aberrations, IEEE Trans. Ultrason. Ferroelect. Freq. Control, vol. 43, pp , January. Larsen, H. G., and Leavitt, S. C. (1980), An image display algorithm for use in real-time sector scanners with digital scan conversion, IEEE Ultrasonics Symposium Proceedings, pp Li, P.-C., and O Donnell, M. (1995), Phase aberration correction on two-dimensional conformal arrays, IEEE Trans. Ultrason. Ferroelec. Freq. Contr., 42: Macovski, A. (1983), Medical Imaging Systems, chap. 10, Prentice-Hall, Englewood Cliffs, N.J. Maslak, S. H., (1979), Acoustic Imaging Apparatus, U.S. Patent 4,140,022, issued February 20. Mucci, R. A. (1984), A comparison of efficient beamforming algorithms, IEEE Trans. Ac., Speech, Sig. Proc., 32: Nock, L., Trahey, G. E., and Smith, S. W. (1988), Phase aberration correction in medical ultrasound using speckle brightness as a quality factor, J. Acoust. Soc. Am., 85: O Donnell, M. O., and Flax, S. W. (1988), Phase aberration measurements in medical ultrasound: human studies, Ultrasonic Imaging, 10:1 11. Park, S. B., and Lee, M. H. (1984), A new scan conversion algorithm for real-time sector scanner, IEEE Ultrasonics Symposium Proceedings, pp: Richard, W. D., and Arthur, R. M. (1994) Real-time ultrasonic scan conversion via linear interpolation of oversampled vectors, Ultrasonic Imaging, 16: Rigby, K. W. (2000), Real-time correction of beamforming time delay errors in abdominal ultrasound imaging, Proc. SPIE, 3982: Sarti, D. A. (1987), Diagnostic Ultrasound Text and Cases, Year Book Medical Publishers, Chicago. Schafer, M. E., Lewin, P. A. (1984), The influence of front-end hardware on digital ultrasonic imaging, IEEE Trans. Ultrason. Ferroelec. Freq. Contr., 31: Schafer, M. E., and Lewin, P. A. (1989), Transducer characterization using the angular spectrum method, J. Acoust. Soc. Amer., 85: Silverstein, S. D. (2001), A Robust Auto-focusing Algorithm for Medical Ultrasound: Consistent Phase References from Scaled Cross-correlation Functions, IEEE Signal Processing Letters, 8(6): Steinberg, B. D. (1976), Principles of Aperture and Array System Design, Wiley, New York. Steinberg, B. D. (1992), Digital Beamforming in Ultrasound, IEEE Trans, on Ultrasonics, Ferro., and Freq. Cont., 39(6): , November.

18 25.18 DESIGN OF MEDICAL DEVICES AND DIAGNOSTIC INSTRUMENTATION Thomenius, K. E. (1996), Evolution of Ultrasound Beamforming, Proc. IEEE Ultrasonics Symposium, IEEE Cat. No. 96CH35993, pp Thurstone, F. L., and von Ramm, O. T. (1973), A new ultrasound imaging technique employing two-dimensional electronic beam steering, vol. 5 in Acoustical Holography, Booth Newell (ed.), pp , Plenum Press, New York. Trahey, G. E., and Smith, S. W. (1988), Properties of acoustical speckle in the presence of phase aberration. Part I: first-order statistics, Ultrasonic Imaging, 10: Wells, P. N. T. (1977), Biomedical Ultrasonics, Academic Press, San Diego. Zhu, Q., and Steinberg, B. D. (1993a), Wavefront amplitude distortion and image sidelobe levels: Part I Theory and computer simulations, IEEE Trans. Ultrason. Ferroelec. Freq. Contr., 40: Zhu, Q., and Steinberg, B. D. (1993b), Wavefront amplitude distortion and image sidelobe levels: Part II In vivo experiments, IEEE Trans. Ultrason. Ferroelec. Freq. Contr., 40: