ultrasonic imaging of the human body - semantic …...ultrasonic imaging of the human body 673 1....

52
Rep. Prog. Phys. 62 (1999) 671–722. Printed in the UK PII: S0034-4885(99)74694-4 Ultrasonic imaging of the human body P N T Wells Department of Medical Physics and Bioengineering and Centre for Physics and Engineering Research in Medicine, Bristol General Hospital, Bristol BS1 6SY, UK Received 30 November 1998 Abstract Ultrasonic imaging is a mature medical technology. It accounts for one in four imaging studies and this proportion is increasing. Wave propagation, beam formation, the Doppler effect and the properties of tissues that affect imaging are discussed. The transducer materials and construction of the probes used in imaging are described, as well as the methods of measuring the ultrasonic field. The history of ultrasonic imaging is briefly reviewed. The pulse–echo technique is used for real-time grey-scale imaging and the factors that limit the spatial and temporal resolutions are considered. The construction and performance of transducer arrays are discussed, together with the associated beam steering and signal processing systems. Speckle and scattering by blood are introduced, particularly in the context of the observation of blood flow by means of the Doppler effect and by time-domain signal processing. Colour flow imaging, and the colour coding schemes used for velocity and power imaging, are explained. The acquisition and display of three-dimensional images are discussed, with particular reference to speed and segmentation. Specialized imaging methods, including endoluminal scanning, synthetic aperture imaging, computed tomography, elasticity imaging, microscanning, contrast agents, and tissue harmonic imaging, are reviewed. There is a discussion of issues relating to safety. Conclusions are drawn and future prospects are considered. 0034-4885/99/050671+52$59.50 © 1999 IOP Publishing Ltd 671

Upload: others

Post on 03-Jun-2020

12 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Rep. Prog. Phys.62 (1999) 671–722. Printed in the UK PII: S0034-4885(99)74694-4

Ultrasonic imaging of the human body

P N T WellsDepartment of Medical Physics and Bioengineering and Centre for Physics and Engineering Research in Medicine,Bristol General Hospital, Bristol BS1 6SY, UK

Received 30 November 1998

Abstract

Ultrasonic imaging is a mature medical technology. It accounts for one in four imagingstudies and this proportion is increasing. Wave propagation, beam formation, the Dopplereffect and the properties of tissues that affect imaging are discussed. The transducer materialsand construction of the probes used in imaging are described, as well as the methods ofmeasuring the ultrasonic field. The history of ultrasonic imaging is briefly reviewed. Thepulse–echo technique is used for real-time grey-scale imaging and the factors that limitthe spatial and temporal resolutions are considered. The construction and performanceof transducer arrays are discussed, together with the associated beam steering and signalprocessing systems. Speckle and scattering by blood are introduced, particularly in thecontext of the observation of blood flow by means of the Doppler effect and by time-domainsignal processing. Colour flow imaging, and the colour coding schemes used for velocity andpower imaging, are explained. The acquisition and display of three-dimensional images arediscussed, with particular reference to speed and segmentation. Specialized imaging methods,including endoluminal scanning, synthetic aperture imaging, computed tomography, elasticityimaging, microscanning, contrast agents, and tissue harmonic imaging, are reviewed. Thereis a discussion of issues relating to safety. Conclusions are drawn and future prospects areconsidered.

0034-4885/99/050671+52$59.50 © 1999 IOP Publishing Ltd 671

Page 2: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

672 P N T Wells

Contents

Page1. Introduction 6732. Physical foundations 673

2.1. Wave propagation 6732.2. Beam formation 6742.3. The Doppler effect 6752.4. Radiation force 6752.5. Basic ultrasonic properties of biological materials 6752.6. The physical dimensions of ultrasonic imaging 6762.7. Tissue inhomogeneity 6762.8. Nonlinear propagation 6772.9. Beam and pulse propagation in real tissue 677

3. Generation and detection of ultrasound 6773.1. Transducer materials 6773.2. Probe construction: elementary considerations 6793.3. Matching, backing and loading: pulse operation 6793.4. Beam studies 681

4. Image formation 6834.1. Principles of pulse–echo ultrasound 6844.2. Transducer array scanning 6874.3. Signal processing and display for grey-scale pulse–echo imaging 6914.4. Resolution 6924.5. Speckle 6934.6. Examples of real-time grey-scale scanning 6944.7. Blood flow and tissue motion imaging 6944.8. Three-dimensional image acquisition and display 7054.9. Specialized imaging methods 708

5. Safety considerations 7176. Conclusions and future prospects 718

Page 3: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 673

1. Introduction

More than one out of every four medical diagnostic imaging studies in the world is nowestimated to be an ultrasound study and the proportion continues to increase (WFUMB 1997).This situation has come about because of the remarkable advances that have taken place in thephysics and engineering of ultrasonic imaging since the medical applications of ultrasonicswere last reviewed, some 30 years ago, inReports on Progress in Physics(Wells 1970). In 1970,the emphasis was on the biological effects, surgical and therapeutic applications of ultrasonics;little more than 20% of that review was concerned with diagnosis. That was a fair balancethen, because medical imaging was dominated by x-radiography, with a small contributionfrom radionuclide studies; neither x-ray computed tomography nor nuclear magnetic resonanceimaging had yet been invented and, except perhaps for applications in obstetrics, gynaecologyand cardiology, ultrasonic imaging was generally regarded only as a laboratory curiosity.

The choice of the best imaging technique in any given clinical situation is based onconsiderations such as the resolution, contrast mechanism, speed, convenience, acceptabilityand safety. Ultrasound scores highly in all of these: spatial resolution of a millimetre can beobtained in abdominal scanning, tissue contrast is good (and can be enhanced by using contrastagents), it is a real-time method, convenient to use, very acceptable to patients, and apparentlyquite safe. It can also be a relatively inexpensive technology. Its main disadvantages are thatits images are spoilt by the presence of bone or gas, and the operator needs a high level of skill,both in image acquisition and interpretation.

Without diagnostic imaging tools, the doctor is blind. Imaging is one of the cornerstonesof diagnosis in modern medical practice: the others are clinical examination and the variousbranches of pathology. Moreover, the applications of imaging are not limited to diagnosis.Increasingly, open access surgery is being replaced by minimally-invasive procedures, andimage-guided intervention is becoming more common. For many of these purposes, ultrasonicimaging is the best method.

2. Physical foundations

2.1. Wave propagation

Ultrasonics is concerned with the propagation in various media of mechanical waves withfrequencies above the range of human hearing which, for convenience, means waves withfrequencies of more than 20 kHz.

Consider an infinitely-small-amplitude plane pressure wave propagating in a perfectlyelastic isotropic medium. The wave equation is

∂2u

∂z2= 1

c2

∂2u

∂t2, (1)

whereu is the particle displacement amplitude,z is the position in space along the direction ofpropagation,t is the time andc is the propagation speed. The speed is related to the elasticityK and the densityρ of the medium in which the wave is travelling, according to the equation

c = (K/ρ)1/2. (2)

At a plane boundary between two media with speedsc1 andc2 respectively,

θi = θt (3)

(sinθi/ sinθt) = c1/c2, (4)

Page 4: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

674 P N T Wells

Figure 1. A simple representation of the ultrasonic beam produced by a disc transducer in ahomogeneous medium.

whereθi , θr andθt are respectively the angles of incidence, reflexion and refraction. At normalincidence,

Ir/Ii = [(Z2 − Z1)/(Z2 +Z1)]2, (5)

whereIi andIr are respectively the intensities of the incident and reflected waves andZ1 andZ2 are the characteristic impedances of the two media. The characteristic impedance of amedium is given by

Z = ρc. (6)

The situation to which equation (5) applies is called specular reflexion and it implies thatthe reflecting boundary is both smooth and extensive in relation to the wavelengthλ. Bydefining a quantityψ related to the size of an obstacle, two situations can be distinguished,each with a corresponding value of scattering cross sectionS:

S = 1 when ψ � λ (7)

S = k4ψ6 when ψ � λ (8)

wherek = 2πf and the frequencyf = c/λ. Thus, specular reflexion is described by equation(7) and Rayleigh scattering (Wells 1977), by equation (8). With obstacles of intermediate size(or with rough surfaces), directional scattering occurs.

2.2. Beam formation

The aperture of the ultrasonic transducer used in medical imaging is usually in the form of acircle or a rectangle. As illustrated in figure 1, the ultrasonic beam can be considered to consistof a near field and a far field. With continuous wave excitation of a disc transducer,

Iz/I0 = sin2{(π/λ)[(a2 + z2)1/2 − z2]}, (9)

whereI0 is the intensity at the surface of the transducer,Iz is the intensity at a distancez fromthe transducer along the central axis of the beam anda is the radius of the transducer. In thefar field, beyond the last axial maximum (atz = a2/λ, provided thata2� λ2), the directivityfunction is

Ds = 2J1(ka sinθ)

ka sinθ, (10)

whereθ is the angle betweenDs and the central axis of the beam andJ1 is the first-order Besselfunction. In the near field, the beam is roughly cylindrical with a series of axial maxima and

Page 5: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 675

minima of decreasing complexity moving away from the transducer. Also, in the near field,the beam can be focused by a lens or other means.

If the transducer is excited to produce a transient disturbance, the ultrasonic pulse has itsenergy spread over a spectrum of frequency, corresponding to its bandwidth. This means thatsingle values cannot be assigned toλ or k in equations (9) and (10). Physically, the beamdiffraction pattern is smeared to an extent which changes during the passage of the pulse.

2.3. The Doppler effect

When an ultrasonic wave is scattered by a target that has a component of velocity alongthe direction of beam propagation, the frequency of the scattered ultrasound is shifted by theDoppler effect. Ifθ is the angle between the direction of target motion and that of the ultrasonicbeam,

v = −fDc/(2f cosθ), (11)

wherev is the speed of the target andfD is the difference between the frequencies of theultrasound transmitted from the transducer and backscattered along the ultrasonic beam,provided thatv � c. The negative sign means that the frequency is shifted downwards ifthe target is moving away from the transducer.

2.4. Radiation force

The energy carried by an ultrasonic wave results in radiation force when the wave directionchanges (e.g., as the result of reflexion) or when ultrasound is absorbed from the wave. Theradiation force,F , resulting from complete absorption is given by

F = P/c, (12)

whereP is the ultrasonic power.

2.5. Basic ultrasonic properties of biological materials

The basic properties of biological materials that are of importance in ultrasonic imaging areattenuation, speed and reflectivity. Typical values for various tissues are given in table 1.Attenuation is due to the combined effects of absorption and, because tissue is inhomogeneousover a range of scales, scattering and reflexion. In soft tissues, absorption is mainly due tospectra of relaxation processes (Wells 1975), which accounts for the nearly linear frequencydependence. In contrast, there is a square law dependence on frequency in simple liquids, inwhich absorption is due to viscosity.

For practical purposes, speed dispersion can be neglected in soft tissues; typically, thespeed increases by about 0.01% MHz−1 (Carstensen and Schwan 1959).

When substituted in equation (5), the values of characteristic impedance given in table 1give an indication of the strongest reflexions likely to occur when an ultrasonic beam travelsthrough the body. There is almost complete reflexion at the boundary between soft tissueand air and also when a beam travelling through soft tissue encounters bone. In contrast, thereflexion at a boundary between different kinds of soft tissues is quite small, leaving most ofthe energy to travel across the boundary and into deeper tissues of the body.

Reflexion at the boundaries between different kinds of tissues is a rather idealized situation.It is the backscattering of ultrasound as it travels through tissues that is generally much morerelevant to the imaging process.

Page 6: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

676 P N T Wells

Table 1. Properties of some materials relevant to ultrasonic imaging. Frequency dependences ofα

apply at least within the range 1–10 MHz (except for bone, for which the range is 1–2 MHz). Datacollected by Duck (1990) and Wells (1977). Gaps in the table indicate that published data are notreadily available.

AttenuationPropagation Characteristic coefficient,α Frequency Nonlinearspeed,c impedance,Z at 1 MHz dependence parameter,

Material (m s−1) (106 kg m−2 s−1) (dB cm−1) of α B/A

Air 330 0.0004 1.2 f 2

Blood 1570 1.61 0.2 f 1,3 6.1Brain 1540 1.58 0.9 f 6.6Fat 1450 1.38 0.6 f 10Liver 1550 1.65 0.9 f 6.8Muscle 1590 1.70 1.5–3.5 f 7.4Skull bone 4000 7.80 13 f 2

Soft tissue 1540 1.63 0.6 f

(mean values)Water 1480 1.48 0.002 f 2 5.2

2.6. The physical dimensions of ultrasonic imaging

Ultrasonic imaging depends on the interactions between the structures of the (human) bodyand ultrasonic radiation. As a starting point, the ultrasonic wavelength can be considered todetermine the spatial resolution. Consider the process of imaging the contents of the abdominalcavity. Structures of interest for imaging are likely to be located at distances of up to about150 mm beyond the skin surface and spatial resolution of the order of a millimetre is needed. Toobtain this resolution, the wavelength of the ultrasound needs to be not greater than 1 mm, whichis the wavelength at a frequency of 1.5 MHz. The problem is that the attenuation increaseswith the frequency, so that the distance over which useful levels of energy can be propagatedbecomes less as the frequency is increased. Typically, 3 MHz might be the maximum frequencyfor 150 mm penetration. The corresponding wavelength is 0.5 mm. Assuming a mean valueof attenuation equal to 0.6 dB cm−1 MHz−1 (see table 1), the total attenuation, for the roundtrip, is about 54 dB.

From equation (9), the aperture required for a disc transducer to produce a beam with anear field depth of 150 mm at 0.5 mm wavelength is 17 mm in diameter; operation in the nearfield is necessary for beam focusing to be effective.

At a frequency of 3 MHz, substitution in equation (11) shows that a Doppler shift frequencyof 4 kHz occurs when ultrasound is reflected by a target moving at a velocity of 1 m s−1 withrespect to the direction of the ultrasonic beam. The fact that the Doppler frequency usuallylies in the audible range in ordinary physiological situations is serendipitous.

2.7. Tissue inhomogeneity

Although it is possible to assign average values to speed, attenuation and scattering in softtissue, the different kinds of soft tissues each have their own individual properties and theseproperties are inhomogeneously distributed within them. For example, table 1 gives a range ofvalues for attenuation in muscle, in which attenuation across the muscle fibres is around twicethat along the fibres.

An ultrasonic beam is distorted as it travels through tissue as a result of tissueinhomogeneity. This is discussed further in sections 2.9 and 4.4.

Page 7: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 677

2.8. Nonlinear propagation

Although it is often convenient to assume that the pressure of a wave is proportional to theparticle displacement amplitude, this is really only justifiable at small amplitudes. Except atinfinitely-small amplitudes, the nonlinear relationship between pressure and density becomes asignificant factor. Physically, nonlinear propagation transfers wave energy to higher harmonics(because propagation speed increases with density) so that an initially sinusoidal wave isconverted to a sawtooth. The resultant shock wave is accompanied by excess attenuation(because attenuation increases with frequency), so that the rate of deposition of wave energyhas a spatial peak at some distance from the source. As the wave amplitude falls, so thewaveform reverts towards a sinusoidal shape.

The nonlinearity of a medium can be described in terms of its nonlinearity parameterB/A.The quantitiesA andB are the coefficients of the first- and second-order terms of the Taylorseries expansion of the equation relating pressure to density in the medium. Typical values ofB/A are given in table 1.

The distribution of energy deposition when a beam propagates in a homogeneous nonlinearmedium is determined by the combined effects of nonlinearity, absorption and diffraction.Aanonsenet al (1984) solved the nonlinear wave equation and, using a modification of theircode, Baker (1997) demonstrated that the location of peak intensity loss from a beam generatedby a circular source shifts away from the source and occupies a smaller diameter as the intensityis increased. Physically, this is because of the build-up of higher harmonic frequencies in thebeam at the intensities at which the effect is significant.

2.9. Beam and pulse propagation in real tissue

Figure 2(a) represents a pulsed ultrasonic beam propagating in an idealized losslesshomogeneous material. The shape of the pulse is unaltered during propagation; its amplitudein the cylindrical near field is constant and only reduces in the far field as the result of beamdivergence.

Real tissue possesses frequency-dependent attenuation and is nonlinear andinhomogeneous. Figure 2(b) represents the propagation of a pulsed ultrasonic beam in tissue.The pulse that leaves the transducer is the same as that represented in figure 2(a). Aftertravelling some distance in real tissue, however, a finite amplitude pulse is not only attenuatedbut has also been partially converted to a shock wave as the result of nonlinearity. Deeperinto the tissue, the pulse has been further attenuated, preferentially at its higher frequencycomponents, and has reverted to a less shocked shape with the remaining energy distributedacross lower frequencies. The beam has also been deviated and distorted by inhomogeneity.

3. Generation and detection of ultrasound

3.1. Transducer materials

The ideal transducer for ultrasonic imaging would be perfectly matched to the (human) body,have high efficiency as a transmitter and high sensitivity as a receiver, a wide dynamic rangeand a wide frequency response for pulse operation.

Transducers based on the piezoelectric effect are almost always used in ultrasonic imaging.Piezoelectricity occurs in many natural materials, quartz perhaps being the best known, butthese natural materials have characteristics which are not very suitable for medical ultrasonics.The artificial ferroelectric ceramics (Jaffeet al 1955) approach much more closely to theideal. When the crystalline structure of a material has no centre of symmetry, it is said to

Page 8: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

678 P N T Wells

Figure 2. Simple representations of the ultrasonic beam produced by pulsed excitation of a disctransducer. (a) The beam in an idealized lossless homogeneous medium. The beam shape is thatillustrated in figure 1. The pulse shape does not change during propagation, buts its amplitudeis reduced in the far field as the result of beam divergence. (b) The beam in an inhomogeneousnonlinear attenuating medium, such as real tissue. The beam is deviated as the result of theinhomogeneity: under some circumstances, there may be several cross sectional maxima. Thepulse develops shocks as the result of the nonlinearity: this leads to rapid attenuation of the higherfrequency components and the centre frequency of the attenuated pulse is shifted downwards.

be noncentrosymmetric. Ferroelectric ceramics have noncentrosymmetric unit cells which,when below the Curie temperature, are randomly orientated, but which can be permanentlypreferentially aligned by the brief application of a large polarizing potential.

The behaviour of a piezoelectric transducer can be described, in terms of its efficiency asa transmitter and sensitivity as a receiver, by its piezoelectric transmittingd and receivinggcoefficients, as follows:

d = gεT (13)

k2ps

E = dg, (14)

whereεT is the free dielectric constant of the material,sE is its elastic compliance at constantfield andkp is its planar coupling coefficient.

In biomedical applications, the lead zirconate titanate (PZT) ceramic ferroelectricmaterials have for many years been the most popular transducers and this continues to bethe case although composites of PZT and plastic polymers are beginning to be used. Thepiezoelectric polymer material, polyvinylidene difluoride (PVDF), has some advantages,particularly as a high frequency receiver.

Table 2 gives values for important electromechanical properties of some materials whichare used as transducers in biomedical applications. Most commonly, transducers are operatedeither at resonant frequency or over a band of frequencies containing the resonant frequency.For fundamental frequency operation of a lead zirconate titanate transducer in the thicknessmode,λ/2 ≈ 2 mm at 1 MHz (and proportionately less at higher frequencies). In summary,the data in table 2 show that PZT4 is a good transducer material for ultrasonic power generation

Page 9: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 679

Table 2. Electromechanical properties of some piezoelectric materials. Data from Bowenet al(1996) and Hadjicostiset al (1984).

Material

Electromechanical property PZT4 PZT5 PVDF PZT/polymer 1-3Coupling coefficient,kp 0.57 0.60 0.20 0.63Charge constant,d33 315 374 35 650(×10−12 m V−1)Voltage constant,g33 24.6 24.8 152 66(×10−3 V m N−1)Dielectric constant,εT 1150 1500 10 460(×10−11 F m−1)Elastic compliance,SE 24 26 133 108(×10−12 m2 N−1)MechanicalQ 600 75 10 20Density,ρ 7600 7700 1400 1800(kg m−3)Wave velocity,c 4000 3760 3000 3330(m s−1)Characteristic impedance,Z 30.4 29.0 4.2 6.0(×106 kg m−2 s−1)Curie temperature,Tc 320 365 130 350(◦C)

but, because of its highQ, it is not as good as PZT5 in pulse operation. PVDF is apparentlya good receiving transducer, but its low dielectric constant and its lowd33 value mean thatits sensitivity is greatly reduced by, for example, connecting cable capacitance, and it isan inefficient transmitter. The great advantage of composite transducer materials, such asPZT/polymer 1-3, is that they have a relatively low characteristic impedance (which is also anadvantage of PVDF), whilst retaining a value of coupling coefficient which actually may evenbe somewhat higher than that of PZT.

3.2. Probe construction: elementary considerations

In medical ultrasound, a device that contains a transducer is often referred to as a ‘probe’,since it is used to generate and detect the ultrasound that probes the internal structures ofthe body. As explained in section 2.2, a simple kind of transducer is in the shape of a disc ofpiezoelectric material: the diameter, in terms of the wavelength, determines the geometry of theultrasonic beam. Unless its behaviour is modified by matching and backing (see section 3.3),such a transducer has its fundamental resonance at the frequency at which its thickness ishalf a wavelength; for PZT, this is about 2 mm at 1 MHz and proportionately less at higherfrequencies. The sharpness of the resonance is described by theQ of the transducer material,so that, for short pulse operation, a lowQ is desirable.

3.3. Matching, backing and loading: pulse operation

The instantaneous particle pressure (p) in a medium supporting the propagation of a wave isgiven by

p = ρcv0 sinωt, (15)

wherev0 is the peak particle velocity,ω = 2πf andt is the time. The derivation of equation (5)depends on the continuities of both particle velocity and particle pressure across the interface

Page 10: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

680 P N T Wells

between two media. It follows from equation (5) that

It/Ii = 4Z2Z1/(Z2 +Z1)2, (16)

whereIt is the intensity of the wave transmitted into the second medium. It is usually the case,in medical ultrasonics, that the transducer has a different (higher) characteristic impedance thanthat of the medium (water or soft tissue) in which the ultrasound is propagated. Substitutionin equation (16) shows that, with a PZT transducer, there is about 20% transmission fromthe transducer into the load. If a lossless layer of thicknessl, characteristic impedanceZ3

and wavelengthλ is included between the transducer and the load, however, the transmissionacross the layer is given by

It/Ii = 4Z2Z1[(Z2 +Z1)2 cos2 kl + (Z1 +Z2Z1/Z3)

2 sin2 kl], (17)

wherek = 2π/λ. Then, if l = nλ/4, wheren is an odd integer,It/Ii = 1 (i.e., there is 100%transmission). Thus, the impedance of the transducer can be matched to that of the load bymeans of aλ/4 layer of a material of intermediate impedance (typically 7× 106 kg m−2 s−1

to match PZT to water). At higher frequencies, a quarter-wavelength layer may be too thinto be reliable, in which case a 3λ/4 layer may be used. A particularly neat solution to theproblem of fabricating a quarter-wavelength matching layer is to integrate the layer with thefront surface of the transducer. Seyed-Bolorforosh (1996) has described how this can be doneby cutting tiny orthogonal grooves on the front of the transducer so that little pillars of PZT ofthe desired height stand proud. After electroding the surface of the transducer at the base of thegrooves, epoxy resin is poured over the structure to fill them. The characteristic impedance ofthe integrated composite material thus formed can be designed to have the appropriate value,by proper choice of the dimensions of the pillars.

For continuous wave operation, a transducer with a highQ is desirable. This means thatits resonant frequency is well-defined and the thickness of the quarter wavelength matchinglayer can be calculated precisely. Because the device operates at resonance, however, it isusually satisfactory for a half-wavelength layer to be used between the transducer and the load.According to equation (17), the characteristic impedance of such a layer does not affect thetransmission; high efficiency is obtained provided that the attenuation in the layer (and in thetransducer) is low. A quarter-wavelength matching layer does have an important advantagewith pulse operation. Because the frequency spectrum of a pulse covers a range that increases asthe pulse length becomes shorter, however, the thickness corresponding to a quarter wavelengthbecomes increasingly ill-defined.

Besides matching the characteristic impedance of the transducer to that of the load, thereare two other factors that have an important influence on the performance of the probe. Thefirst is the backing, or rear surface loading, of the transducer. A simple approach, for shortpulse operation, is to back the transducer with a medium similar in characteristic impedance tothat of the transducer, and which absorbs any ultrasound that leaves the transducer from its rearsurface (Nguyenet al 1996). This has the effect of damping the resonance of the transducerso that it is equally sensitive over a wide range of frequency. The penalty is that the sensitivityis very significantly reduced. Good performance is obtained with impedance matching ofthe front surface of the transducer with a quarter-wavelength plate and with an air-backedrear surface, or, better still, with a rear half-wavelength plate tuned to a frequency which,in conjunction with the resonant frequency of the front plate, extends the overall frequencyresponse of the device.

The other factor that influences the performance of the probe is the electrical circuit towhich it is connected. At resonant frequency, the transducer behaves like an inductorL anda capacitorC connected in series (with a resistorR, also in series, to account for loss in thedevice), with a second capacitorCs connected in parallel. The series elements correspond to

Page 11: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 681

the mechanical behaviour of the device, the resonant frequencyω0 = 2πf0 being determinedby the equationω0 = 1/(LC)1/2. Also,Q = ω0L = 1/ω0CR. The effect of the capacitorCs is to reduce the sensitivity of the device. By connecting an inductorLs in series with thetransducer, the total impedance of the reactive components loading the transducer becomesequal to zero at resonant frequency, i.e., whenω0 = 1/(LsCs)

1/2.It can now be seen that the probe designer can select the characteristics of the matching

and backing layers and the electrical loading of the transducer to optimize the performanceof the device. Generally, this means that the sensitivity is constant and has its maximumpracticable value over the desired frequency bandwidth. Such a device is also likely to havethe maximum practicable dynamic range: this means that the electrical ripple that immediatelyfollows the transmission or reception of a brief pulse of ultrasound has the minimum amplitudeand duration.

3.4. Beam studies

The pressure distribution of the ultrasonic beam produced by a transducer with any sizeand shape of aperture, and with any temporal distribution, can be theoretically calculated.For example, equations (9) and (10) describe the beam produced by a disc transducer withcontinuous wave excitation. In practice, however, insufficient data are likely to be available toallow a reliable estimate of beam shape to be calculated and it is necessary to measure or toobserve the spatial distribution of, for example, the pressure field. In addition, the integratedbeam power and the intensity distribution may be of interest. The most important of theavailable measurement and observation methods are described in the following sub-sections.

3.4.1. Hydrophones. Hydrophones are detectors based on transducers that respond directlyto the ultrasonic field. The output of a hyrophone is an electrical signal that follows theinstantaneous value of the ultrasonic pressure, ideally over a small area.

A small area hydrophone needs to be small in relation to the wavelength of the ultrasoundthat it is required to measure. This makes the device nondirectional. There are two main typesof construction. A needle or fibre hydrophone has a sensitive tip and can be used to probean ultrasonic field. A membrane hydrophone consists of a relatively large sheet of a thin filmof plastic piezoelectric material that has a negligible effect on the field; a tiny element of themembrane is electroded to form a sensitive area.

A typical piezoelectric needle hydrophone consists of a polyvinylidene difluoride disc,500µm in diameter and 15µm in thickness, attached to an insulating layer at the tip of a600µm diameter stainless steel tube (Lewin 1981). The needle lumen contains a backingmaterial that has a higher characteristic impedance than water, so that resonance occurs at thefrequency at which the transducer isλ/4 in thickness.

There are two main types of fibre-optic hydrophones. In one type, the ultrasonic pressurewave modulates the refractive index of the fluid in front of the tip of the fibre, leading to achange in reflectivity that can be measured with an optical detection system at the other endof the fibre (Staudenraus and Eisenmenger 1993). A problem with this type of hydrophone isits lack of sensitivity. The second type of fibre-optic hydrophone has a metal-coated tip andoperates as a baseband or heterodyne interferometer. In a variant of this approach, a robustdevice, capable of withstanding shock waves, consists of a cut fibre with one or more harddielectric optical layers formed by sputtering. For example, a single-mode fibre with a corediameter of 3.5µm with a λ/4 layer at the tip has been constructed, with a bandwidth of70 MHz (Koch 1996).

A typical membrane hydrophone consists of an annular frame with a diameter of 100 mm,

Page 12: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

682 P N T Wells

over which a 9µm-thick sheet of polyvinylidene difluoride is stretched (Prestonet al 1983).This thickness corresponds toλ/2 at 170 MHz. At the centre of the disc of film, a sensitiveelement is formed by vacuum-deposited gold electrodes with diameters in the range 0.2–1.0 mm, with nonoverlapping connexion tracks. The membrane essentially has a negligibleeffect on an ultrasonic field in the 1–15 MHz frequency range and the sensitivity is flat within1 or 2 dB.

3.4.2. Target plotting. A method that is quite widely used for plotting the effective distributionof an ultrasonic beam, provided that it can be pulsed to allow echoes to be detected, is to scana small target point-by-point within the field. For echo amplitude measurement, a null methodcan be used with a calibrated attenuator appropriately positioned in the electrical signal path.

The choice of target has to recognize the compromise between the desirability of small size(to minimize directionality) and the need for the echo to be detectable. Also, the reflectivityof the target should vary with the ultrasonic frequency in a well-defined fashion. Lypacewiczand Hill (1974) concluded that a small stainless steel ball bearing is an excellent choice: it isself-aligning and nondirectional, and can be supported by a wire soldered to its rear surface.

3.4.3. Schlieren observation.The variation in the density of a medium supporting anultrasonic wave produces corresponding changes in the optical refractive index if the mediumis transparent. The schlieren method depends on this phenomenon. A parallel beam of lightis arranged to pass through a transparent medium (usually water) in which the ultrasonicbeam is travelling. The light is then focused onto an obstruction, so that none reaches theobserver when there is no ultrasonic field. When ultrasound changes the refractive index ofthe medium, however, some of the light that passes through the disturbed area no longer fallson the obstruction; that which is deviated constitutes a bright image of the ultrasonic beam.Thus, the beam becomes visible against the dark background. A closed-circuit TV system canbe conveniently used for display.

Pulsing the light synchronously with the ultrasonic pulse generator results in stroboscopicvisualization of the field. If viewed via a TV camera, an indication of the beam profile canbe obtained by displaying the amplitude-time waveform of an appropriately selected TV scanline (Follett 1986).

3.4.4. Magnetic resonance imaging.Nuclear magnetic resonance signals can be influencedby motion within the material being studied by the application of a magnetic field gradientsuperimposed on the static, uniform magnetic field that is used for spin polarization (Hahn1960). Particle displacements of a few micrometres accompany the propagating wavestypically used for ultrasonic imaging. Ultrasonic waves in agar gel have been visualizedby a clinical MRI scanner with synchronized phase-locked gradient waveforms (Walkeret al1998). The method has the potential to enable the propagation of ultrasonic beams used forimaging to be studied noninvasively in living tissues.

3.4.5. Power and intensity measurement.There are three main methods for measuringthe power transported in an ultrasonic beam. These are by calorimetry, radiation forcemeasurement, and calibrated hydrophone.

In calorimetry, the quantity of heat produced in unit time when an ultrasonic beam iscompletely absorbed is equated to the ultrasonic beam power. Essentially, there are twotypes of calorimeter. In a flow calorimeter, a liquid is continuously pumped through thevessel in which the ultrasound is absorbed. The power is calculated from the difference in

Page 13: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 683

the temperatures of the outflow and inflow liquid, and the rate of flow of the liquid and itsspecific heat. A convenient improvement is to place an electrical heating element betweenthe temperature measurement sites. The reduction in electrical power needed to maintaina constant temperature difference when the ultrasound is switched on is then equal to theultrasonic power. This is independent of the flow rate and the specific heat, and the electricalpower can be controlled by a feedback loop. This type of calorimeter is relatively sensitiveand has a fast response (Torr and Watmough 1977).

In the second type of calorimeter, flowing liquid is not used to maintain a steady state.Instead, the rate of rise of temperature of the device is measured whilst the ultrasound is beingabsorbed. The water equivalent of the device needs to be known, for the ultrasonic power tobe calculated, or it must be calibrated against another power measurement method. Althoughthe initial rate of rise of temperature is only slightly affected by cooling to the environment, acooling correction needs to be applied if the heating leads to a significant temperature gradient.An example of a calorimeter of this type, consisting of a hollow sphere filled with carbontetrachloride (the characteristic impedance of which is close to that of water, and which isrelatively absorbent) and containing an array of thermocouples has been described by Wellset al (1963).

The importance of ensuring that the calorimeter responds only to heat generated by theabsorption of ultrasound, and not to that resulting from the inefficiency of the transducer, doesnot seem to have been widely realized. Effective thermal isolation is necessary.

The relationship between the radiation force produced when an ultrasonic beam travellingat a known velocity is absorbed, and the ultrasonic power transported by the beam, is givenby equation (12). For an ultrasonic wave with a power of 1 W travelling in water, a force of670µN is produced when the wave is completely absorbed. This is equivalent to the forceof gravity acting on a mass of 69 mg. The force is doubled if the beam is totally reflected atnormal incidence. Thus, the ultrasonic powers used in imaging can be measured by water-immersed instruments based on, for example, modified analytical balances (Rooney 1973) orelectronically-controlled servo sensors (Farmery and Whittingham 1978).

There are several approaches to the measurement of ultrasonic intensity. For example, acalibrated hydrophone may be used. The signal detected by a hydrophone is proportional to theparticle displacement amplitude. The intensity of the ultrasound is proportional to the squareof the displacement amplitude. Provided that the total beam power is known (for example, byradiation force measurement), scanning the hydrophone point-by-point across the beam andintegrating the squares of the resultant voltage measurements provides the data necessary tocalibrate the hydrophone. Other methods of intensity measurement include the radiation forceon a small suspended spherical target (Hasegawa and Yoshioka 1969), the temperature riseof a small thermocouple junction embedded in an absorbent disc (Fry and Fry 1954), and thedisplacement of a thin reflective pellicle observed by laser interferometry (Vilkomersonet al1977). Generally, methods that use piezoelectric or optical detectors are likely to be moresensitive, and to have better time resolution, than those using thermal detectors. Therefore,they are more likely to be suitable for measuring the relatively low time-averaged intensitiesthat are usually used in ultrasonic imaging.

4. Image formation

The first attempts to use ultrasound for medical diagnosis were based on the expectation thatit would be possible to demonstrate tissue masses within the body and, particularly, withinthe brain, because of differences in attenuation. Dussiket al (1947) constructed a scannerin which a beam of ultrasound was directed through the patient’s head and detected by a

Page 14: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

684 P N T Wells

receiver placed in line with the transmitter. The images which were formed by scanningthe beam in a raster pattern seemed to represent the intracerebral structures, including theventricles. Using a similar scanner operating at a frequency of 2.5 MHz and an intensity ofabout 1 W cm−2, Hueter and Bolt (1951) concluded that ‘a preliminary evaluation indicatesthat the echo-reflection method is considerably less promising (than the transmission method)for general ventriculography, mainly because of the small amount of reflection at the interfacebetween the tissue and the ventricular fluid’. The subsequent demonstration (Ballantineet al1954) that an empty skull gave rise to similar pictures, because of the coincidental transmissionproperties of the bone, halted work on this approach and arguably held back progress inultrasonic imaging research for several years.

Pulse–echo ultrasound was shown to have practical value for the detection of flaws inmetals during World War II; the publications of Firestone (1946) in the USA and Deschet al(1946) in the UK appeared as soon as the constraints of military secrecy were relaxed. Usingthis same technique, research into medical applications soon began in Denver, where Howryand Bliss (1952) constructed a water-immersion two-dimensional scanner, and in Minneapolis,where Wild and Reid (1952) started to develop high-frequency real-time two-dimensionalimaging. From this early work, researchers world-wide and in increasing numbers began toexplore the potential of the new technique, although perhaps initially more slowly in the USAthan elsewhere, because of the set-back with transmission imaging.

4.1. Principles of pulse–echo ultrasound

The pulse–echo method depends on the measurement of the time that elapses between thetransmission of a pulse of ultrasound and the reception of its echo from a reflecting or scatteringtarget (from which the distance to the source of the echo can be calculated, if the propagationspeed is known), and the measurement of the amplitude of the echo (which is related to theultrasonic properties of the target). The spatial resolution is determined, in elevation and inazimuth, by the cross sectional dimensions of the ultrasonic beam, and, in depth, by the durationof the ultrasonic pulse. The maximum depth of penetration is that at which the amplitudesof echoes are just detectable; this depends on the attenuation of ultrasound in the tissue,which is itself dependent on the ultrasonic frequency. Ultimately, the resolution is limited bydiffraction. This means that spatial resolution improves as the wavelength is decreased (i.e.,as the frequency is increased), but this has to be set against the consequential reduction inpenetration.

4.1.1. Line scanning. Ultrasonic pulse–echo information is generally acquired along anultrasonic beam and, whilst this is in progress, the beam has effectively to be in a fixedspatial position. Thus, the pulse–echo wavetrain is a single line of information in whichtime corresponds to depth and amplitude, to the reflectivity, or backscattering strength, of thetissue structures along the beam. Displayed on, for example, a cathode ray oscilloscope, thewavetrain is called an ‘A-scan’, using terminology originating in radar.

The essential components of an ultrasonic A-scope are illustrated in figure 3. In a typicalarrangement designed to penetrate 150 mm into the body, the ultrasonic centre frequency couldbe 3 MHz and the transducer could have a diameter of 17 mm (see section 2.6). Range ambiguitywould arise with pulse repetition frequencies in excess of 5000 Hz (because ultrasound travels300 mm in soft tissue in 200µs). The pulse repetition frequency is determined by the rategenerator and, in practice, a frequency of 200 Hz would typically be used as this is high enoughto avoid display flicker and it seems prudent to minimize the exposure of tissue to ultrasound.Echoes from deeper structures are increasingly attenuated by the intervening tissue; the swept

Page 15: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 685

Figure 3. Basic elements of the A-scope. The output from the receiver is connected to the vertical(y) deflexion plates of the cathode ray tube, and that from the timebase generator, to the horizontal(x) plates.

gain generator increases the amplification of the receiver, following the transmission of eachultrasonic pulse, to compensate for this.

4.1.2. Two-dimensional scanning.The production of an image of a plane section throughthe body can be accomplished by scanning the ultrasonic beam across the plane whilst pulse–echo wavetrains are acquired. A two-dimensional image is formed by relating the positionsof registrations on the display to the positions of the corresponding echo-producing structureswithin the patient, as illustrated in figure 4. This requires a means to determine the ultrasonicbeam position in the scan plane so that the two timebases of the display, one horizontal andone vertical, produce a resultant image line with the appropriate position and orientation. Thebrightness along this line is controlled by the amplitude of the corresponding echo signal.

Various methods can be used to control the direction of the ultrasonic beam and to scanit through the patient. The first type of two-dimensional scanner, that came into widespreadclinical use in the mid-1960s, used a system of sliding or rotating linkages to constrain theultrasonic probe within a fixed plane, the orientation of which could be selected accordingto the anatomical section to be imaged (Donaldet al 1958, Holmeset al 1965, Wells 1966).The configuration that turned out to be most popular was that devised by Wells (1966) andused two articulated arms with a system of wires, pulleys and potentiometers to measure theposition and direction of the ultrasonic beam within the scan plane. The probe was moved byhand across the patient’s skin, with a water-based gel to exclude air; the process of acquiringa single image typically occupied 5–30 s. The scanners were generally designed to producetissue maps with emphasis on the display of organ boundaries. The best scans were consideredto be those in which the anatomy was depicted by thin white lines on a black background, for

Page 16: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

686 P N T Wells

Figure 4. The principles of two-dimensional B-scanning. The timebase line on the display isthe resultant of horizontal and vertical timebases controlled by the scanner to correspond to theorientation and position of the ultrasonic beam in the patient. The amplitudes of the received echoescontrol the brightness of the display.

which purpose a black-and-white bistable display is ideal. The most accomplished operatorsdeveloped the skill to oscillate the probe through an angle whilst moving it around the body(a process known as ‘compound scanning’), thus improving the chances of achieving normalincidence with organ boundaries and so obtaining the strong echoes necessary to form goodimages with the insensitive equipment that was then available.

In the mid-1970s, however, mainly because of the work of Kossoff (1972), it was realizedthat the echo amplitude conveys useful diagnostic information and that grey-scale displays aregenerally greatly superior to those limited to presenting black-and-white images. Nowadays,grey-scale imaging is universally used. The grey-scale capability can be described in termsof the dynamic range. This can be expressed as the separation (in dB) between the minimumand maximum echo amplitudes over which changes in echo amplitude produce perceptiblechanges in image brightness.

The introduction of grey-scale scanning was accelerated by the arrival of the scanconversion memory tube. In this device, the image formed by the intensity-modulated timebasecorresponding to the ultrasonic beam position in the patient is stored as a charge pattern on aninsulated target within a cathode ray tube. The charge pattern is then read out by raster-scanningthe target in a TV-compatible format. This made it possible to replace the old-fashionedphotographic recording methods, and the bistable storage tube, with a convenient method ofgrey-scale display. Previously sceptical doctors began to appreciate the potential of ultrasoundto provide clinically useful images that, with a little experience, could be understood by all.

4.1.3. Real-time scanning.The next major advance in clinical ultrasonic imaging came withthe development of real-time scanning. With a few exceptions, the scanners that were in clinicaluse up to the mid-1980s required at least a few seconds to acquire an image. Apart from the

Page 17: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 687

limitation that motion, such as that of the heart, could not be directly observed, the processof ultrasonic diagnosis consisted of making a scan, examining the image, deciding whetheranother scan was needed, and so on until there was reasonable confidence in the result. Theprocess was revolutionized by the development of scanners that acquired grey-scale imageswith frame rates of 15–20 s−1 and higher, producing flicker-free displays in real time, so that theimage immediately followed changes in scan plane orientation and physiological movementcould be observed.

The maximum pulse repetition rate for unambiguous pulse–echo signal acquisitiondepends on the required depth of penetration into the patient. For example, it is 5000 Hz fora maximum penetration of 150 mm (see section 4.1.1). This means that 5000 lines of imageinformation can be acquired per second. Provided that the beam can be scanned through thetissue plane with sufficient speed, the product of the number of lines per frame and the imageframe rate is equal to the pulse repetition rate. Thus, in this example, real-time images eachconsisting of 250 lines could be acquired at a frame rate of 20 s−1, and so on.

The first practicable real-time scanners employed mechanical means to sweep theultrasonic beam through the tissue plane. In one system, two transducers were mountedopposite each other on the rim of a rotating wheel, to produce radial beams; the wheel was atthe focus of a parabolic mirror in a water bath, so that a scan with a rectangular format couldbe made through a flexible membrane forming one wall of the water bath, facing the mirrorand in contact with the patient’s skin (Patzoldet al 1970). Although it was cumbersome touse and sometimes hard to obtain the desired scan plane, this scanner had the advantage ofnot needing a scan converter (because the ultrasonic beam positions were all in parallel, sothe vertical timebase of the display needed only to be translated horizontally to provide imageregistration). Being commercially available and with virtually no competition at the outset, itled the field for several years.

Real-time scanning with a hand-held probe became possible with the introduction ofsystems employing either an oscillating transducer scanning through an oil layer behind athin membrane in contact with the skin (McDickenet al 1974) or in direct contact withthe skin (Eggletonet al 1975, Schuetteet al 1978), or a continuously-rotating wheel withradially-mounted transducers, first in contact with the skin (Holmet al 1975) and, later, whencommercially available, within a liquid-filled casing. All these types of scanners steered theultrasonic beam through a sector and, although it was feasible, in principle, to generate theappropriate timebase control directly, it turned out to be easier and more convenient to convertthe scan to a TV format. This was because digital scan converters with sufficient dynamic rangeand resolution, and based on solid-state random access memories, had just become available ateconomic prices. Nowadays these devices are universally used for scan conversion and imagestorage, except in the very least expensive scanners.

4.2. Transducer array scanning

The methods of two-dimensional scanning that have so far been described all employ single-element transducers. What this means is that the transducer, usually in the form of a disc, hasan aperture that is large enough to produce a directional beam with a near field long enoughto allow focusing to be used to optimize the resolution in azimuth and elevation. The beamsteering is carried out mechanically.

It was probably Buschman (1965) who first used an array of transducers to producean ultrasonic image. His probe had ten small transducers mounted on an arc-shaped supportdesigned to fit over the eye. The transducers were activated sequentially to produce a scan withten discrete lines of image information. Then, well ahead of his time, Somer (1968) described

Page 18: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

688 P N T Wells

the first phased array real-time two-dimensional sector scanner. This was a remarkableachievement. A 10 mm× 11 mm, 1.3 MHz, 21 element transducer array was constructed,together with the electronic circuitry required to steer the transmitted beam through a sector byintroducing the appropriate time delays in the impulses applied to each element in the array.In this first instrument, there was no provision to steer the received beam. The idea was touse the central transducer element in the array as a nondirectional receiver. Somer (1968)was working with neurologists who wanted to produce images of the brain through the intactskull. It was probably mainly because of the unfavourable ultrasonic properties of the skull(reminiscent of the problems explained in the early work described at the beginning of thissection) that the results were disappointing.

In what can now be seen to be a natural extension of Buschman’s (1965) array oftransducers, Bomet al (1971, 1973) constructed linear arrays with 20 transducer elements.Overall, the probe face was 80 mm long and 10 mm wide. Operating at a nominal frequencyof 3 or 4.5 MHz according to the construction of the probe, each element had a diameter of3 mm. The length of the near field of the beam of each element was only 4.5 mm at 3 MHz, andthe half-angle of divergence was 12◦, so the imaging characteristics were far from satisfactory.They were the best that were practicable, however, because the compromise was betweenresolution and image line density: an image with only 20 lines was considered to be as sparseas could be tolerated. The advantages of the system were the simplicity of the display (eachelement acquired a pulse–echo wavetrain in sequence, and the horizontal timebase was shiftedvertically according to the element that was activated) and its rapid frame rate (150 s−1). Theinstrument was vigorously promoted commercially and the potential of real-time imaging torevolutionize cardiological diagnosis became apparent through its use. Only a small numberof enthusiasts acquired instruments, however, because the system was inadequate for routineclinical use.

This type of array of single-element transducers is really quite different from what isnowadays understood by a transducer array. A modern transducer array allows the size of theaperture to be selected and the ultrasonic beam to be focused and steered. Figure 5 shows theprinciples of beam focusing and steering with an array. In this diagram, only four elementsare represented, to simplify the description: a real aperture would typically have at least 16elements. The important point is that each individual element is narrow enough effectivelyto be nondirectional in the scan plane. As viewed in the diagram, each transducer elementemits a cylindrical wavelet in response to electrical excitation. When all the elements in theaperture are excited simultaneously (figure 5(a)), the wavelets combine to form a wavefrontparallel to the aperture, so that the beam travels directly away from the array. With a lineartiming excitation gradient across the array (figure 5(b)), however, the beam is steered in adirection the angle of which to the normal can be changed by changing the direction and slopeof the gradient. In figure 5(c), the situation that arises when the distribution of the timing atexcitation across the array is cylindrical, is that the corresponding ultrasonic wavefront is alsocylindrical: this means that the beam is brought to a focus at the centre of the cylinder. Finally,both the direction and focal length of the beam can be controlled by simultaneously changingboth the gradient and cylindrical radius of the excitation (figure 5(d)).

This explanation of beam control is in the context of the formation of a transmitted beam.The same control of a received beam can be provided by delay lines in the individual signalpaths associated with each element in the array, prior to summing the signals that have passedthrough the delay lines.

Figure 6 shows a linear array consisting of a large number of tiny elements. A typicalmodern array might have 128 elements, each 0.5 mm wide and 7.5 mm long, extending overa distance of 75 mm. In this example, an aperture with a width of 10 mm can be formed by

Page 19: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 689

Figure 5. Principles of beam forming and steering with a transducer consisting of an array of narrowelements, illustrated with four such elements. The same principles apply both on transmission andon reception. (a) Simultaneous excitation produces a beam normal to the array. (b) Linear time-graded excitation steers the beam away from the normal. (c) Cylindrical time-graded excitationfocuses the beam. (d) Superimposed linear and cylindrical time-graded excitation steers and focusesthe beam.

utilising 17 elements in a group. By stepping along the array (Whittingham 1976), one elementat a time, there are 111 discrete beam positions along the array. Even more beam positionscan be formed, if the aperture size is altered alternately by one or two elements during thestepping process. Beam focusing in elevation is provided by the cylindrical lens. (Note thatsuitable lens materials (i.e., materials that have characteristic impedances similar to that ofwater or tissue, but different propagation speeds) usually have higher propagation speeds thanwater or tissue. Consequently, a focusing transducer has a concave section.) This means thatfocusing in elevation has to be at a fixed depth, both on transmission and reception. The sameapplies to focusing in azimuth of the transmitted beam: the focusing conditions cannot be

Page 20: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

690 P N T Wells

Figure 6. Linear transducer array with electronically-controlled focusing in azimuth and lensfocusing in elevation. In this example, there are 12 elements in the active aperture (shown bystippling); there are 51 elements in the array, giving 42 separate lines in the image.

changed after the beam has left the aperture. To give an idea of the magnitude of the delayrequired to focus a beam, a simple trigonometrical calculation shows that the outermost limitof a 10 mm wide aperture has to be excited 170 ns before the centre of the aperture in order tofocus the beam at a depth of 50 mm. On reception, however, the focal length in azimuth can beswept continuously to coincide with the instantaneous position of the echo-producing targets,by dynamically adjusting the delays in the received signal paths from each of the transducerelements within the aperture. The time delays may be introduced either by analogue or bydigital circuits. Although analogue delay lines have largely been superseded by digital circuits,it is of interest to note that three different analogue approaches have been adopted. In oneapproach, the signal is transmitted along a tapped coaxial cable, and a switch extracts thesignal from the tap that happens to provide the desired delay. Alternatively, the delay can beprovided by a tapped series of inductor–capacitor elements. Finally, the capacitative elementin a single inductor–capacitor circuit can be under voltage control, for example, by changingthe depletion layer thickness in a semiconductor diode.

Analogue delay lines have largely been replaced by digital techniques, now that fastsampling and processing speeds with sufficient dynamic range are readily available. At anominal frequency of 5 MHz, for example, digital sampling at a frequency of 50 MHz isadequate to process a typical ultrasonic pulse. After analogue preprocessing of the receivedecho signals, a dynamic range of 50 dB is likely to be sufficient. Digitization to 8 bitscorresponds to 256 levels or 48 dB. The signal having been satisfactorily digitized at 50 MHz,a time shift of 20 ns can be introduced by shifting the waveform by one sample period. Thisis generally adequate for beam focusing and steering.

Page 21: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 691

Figure 7. Examples of typical probes employingtransducer arrays. Left-to-right: a phased array probefor sector scanning; an endovaginal probe; a largecurvilinear array for general purpose abdominal scanning;and a smaller curvilinear array for scanning relativelysuperficial structures. The sizes can be gauged from the15 mm diameter of the endovaginal probe.

The ultrasonic beam from an array can be steered through an angle by introducing a lineartime gradient along the aperture. For example, to steer the beam from a 10 mm aperturethrough an angle of 45◦, the time difference across the aperture is 4.7µs. If the beam isboth to be focused and steered, a cylindrical timing profile needs to be superimposed on thelinear gradient. Although a linear array can be operated in this way, usually to produce ascan with the format of a parallelogram, beam steering is more commonly used over the entireaperture of what is commonly called a ‘phased array’. This produces a sector scan format.A phased array typically has 64 elements in a 15 mm aperture, each element being 200µmwide and 10 mm long. The manufacture of such an array, with a separate electrical connexionto each element and, usually, multiplexing electronic circuits within the probe casing, and thespecialized multicoaxial connecting cable, requires a high level of precision engineering.

The characteristic impedance of a ceramic transducer is substantially different from that ofwater or soft tissues (see table 2). This reduces the sensitivity of the transducer (i.e., increasesits insertion loss). A quarter-wave matching layer is used (see section 3.3) to improve thesensitivity of the system and, together with a matching layer on the rear surface of the transducer,the characteristics can be tuned to provide optimal sensitivity over a wide frequency band.

Examples of typical probes employing transducer arrays are shown in figure 7.

4.3. Signal processing and display for grey-scale pulse–echo imaging

The essential components of the signal processing chain for array scanning are shown infigure 8. The ultrasonic pulse is usually generated by applying a brief (typically 10 ns)monopolar signal, of around 100 V in amplitude, to the transducer (or to each transducerelement in the array). The nominal frequency of the ultrasound is determined by the resonanceof the transducer and its impedance matching layers. In figure 8, the ultrasonic beam issteered and focused, both on transmission (by the multiple time-controlled transmitters) and

Page 22: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

692 P N T Wells

Figure 8. The essential components of the signal processing chain for scanning with a transducerarray.

on reception (by the multiple controlled delay lines), as determined by the beam former. Thevoltage produced by the transducer, in response to echoes from tissues within the patient, istypically in the range 500 mV to 50µV; i.e., it covers a dynamic range of 80 dB. The nonlinearpreamplifiers compress this dynamic range to 50 dB, so that the amplitudes of the signals fedto the multiple controlled delay lines (whether analogue or digital) cover the voltage range1 V to about 3 mV. The swept gain amplifier is able to compensate for substantially more thana range of 50 dB of tissue attenuation (because of the action of the nonlinear preamplifiers),so the voltage range fed to the demodulator is around 5 V to 150 mV.Although the sweptgain time function is usually selected by the operator, an ingenious adaptive technique may beused that is based on the assumption that local attenuation is correlated with local backscatter(Hughes and Duck 1997). The pulse-shaping video amplifier has a frequency response thatoptimizes the appearance of the image and produces an output voltage suited to the type ofdisplay (e.g., cathode ray tube or liquid crystal).

4.4. Resolution

The spatial resolution depends on the profiles of the ultrasonic beam and pulse (see section 3.4)and the characteristics of the signal processing and display system (see section 4.3). Typically,the spatial resolution in elevation is likely to be about three times worse than in azimuth. Atransducer array can be used either as a simple aperture, in which all the elements are activethroughout the signal acquisition process, or the size of the aperture can be adjusted to optimizethe spatial resolution throughout the depth of penetration. In designing an imaging system,an obvious criterion is to maintain a constantf -number (focal length/diameter), independentof the axial position of the target. This can be achieved by expanding the effective size of the

Page 23: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 693

aperture (i.e., by increasing the number of elements active in the array) with time followingthe transmission of the pulse, so that the receiving beam width at the focus remains constant(Harriset al 1991). In addition to this, there are several other advantages with this approach.Because of tissue inhomogeneity (see section 2.7), there is a maximum aperture size beyondwhich there is no further improvement in spatial resolution; indeed, the resolution may actuallybecome worse (Mosfeghi and Waag 1988). Although techniques have been tried to compensatefor the effects of tissue inhomogeneity with both one-dimensional (Smithet al1986) and two-dimensional (Ries and Smith 1995) arrays, none has proved to be easily implementable.

The temporal resolution of an ultrasonic imaging system is limited by the rate at whichimage frames of adequate quality can be acquired. The discussion presented in section 4.1.3applies to the situation in which only one ultrasonic beam is active at any particular time. Theimage frame rate can be increased, however, by parallel processing, although usually at theexpense of some image degradation. For example, Shattucket al (1984) used a broadenedtransmitted beam within which four narrow received beams could be simultaneously formed.In this case, the cost was the reduction in both sensitivity and spatial resolution. In principle,it would be possible, if the array were sufficiently long, simultaneously to employ two entirelyseparate pulse–echo beams; the cost would be an increase in noise due to crosstalk, whichwould be evident as a reduction in image contrast resolution.

The contrast resolution of an ultrasonic imaging system is a measure of its ability to registerperceptibly different display brightnesses from targets of minimally differing reflectivities. Ina perfect imaging system, the brightness transfer characteristic could be selected to allow eventhe minimal change in target reflectivity to produce a perceptible brightness change. Thedisplay of a real imaging system contains ‘noise’, however, including that generated in itselectronic circuits. This noise reduces the contrast resolution. Another important source ofnoise is due to the finite size of the ultrasonic beam and pulse and, particularly, to the sidelobes of the beam which are due to the finite size of the aperture and the grating lobes whichaccompany a beam formed by an array of transducer elements (von Ramm and Smith 1983).For a given number of elements in the aperture, the grating lobes tend to increase in amplitudeas the beam is steered away from the perpendicular direction. The effect of the finite beamwidth and the ancillary lobes is that echo-producing targets lying away from the central axisof the beam may give rise to signals that reduce the contrast resolution of the image.

4.5. Speckle

In pulse–echo ultrasonic imaging, it is the backscattered waves that provide the diagnosticinformation. Although in reality the situation is complicated, blood is a tissue that can beconsidered to consist of isotropic Rayleigh scatterers (see section 2.1). The backscatteringincreases with the fourth power of the frequency. Of course, attenuation in intervening tissuealso increases with frequency, so the frequency that gives the maximum echo amplitude fromblood is determined by the combination of these two effects (Reid and Baker 1971).

Two-dimensional ultrasonic images of blood and the fine structure of soft tissues areactually speckle patterns (Wells and Halliwell 1981, Wagneret al 1983). The ultrasonic pulseoccupies a volume of tissue that contains some number of individual scatterers of varyingstrength and position; the amplitude of the corresponding electrical echo signal from thereceiving transducer is the result of interference between the scattered waves, each of which hasits own particular phase angle (O’Donnell 1983, Finette 1987). In many tissues, the scatteringis primarily due to collagen. Although the correlation function for tissue has not yet beendetermined, the Gaussian model of scattering has so far provided a consistent description oftissue structure (Insanaet al 1990). Initially, as the frequency is increased, the backscattered

Page 24: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

694 P N T Wells

echo amplitude increases towards a maximum value but, at higher frequencies, increasingattenuation in the intervening tissue dominates and the echo amplitude falls.

Because it does not have a one-to-one correspondence with scatterers in the tissue, speckleis sometimes thought merely to be an annoying image artifact, in the same category as noise.More importantly, speckle reduces target detectability; strictly speaking, it does not affectcontrast resolution although it does reduce the useful spatial resolution. Speckle can bereduced by summing uncorrelated images of the scan plane. The requisite uncorrelated imagescan be obtained either by scanning from several different directions, by scanning at severaldifferent ultrasonic frequencies (Gehlbach and Sommer 1987), or by collecting scans froma fixed position while small physiological movements result in differing speckle patterns inimages with essentially the same anatomical information (Wells and Halliwell 1981). In somesituations, none of these methods may be practicable. If this is the case, speckle may bereduced by adaptively filtering the image in the signal processing circuits (Bamber and Daft1986, Chenet al 1996).

Although it might be supposed that speckle suppression would lead to an improvement inimage perception, this may not necessarily be so. The speckle in ultrasonic images is not fullydeveloped and its texture is influenced by larger-than-Rayleigh scatterers. For this reason,the image textures from different tissues may have differing appearances, which can assist inclinical image interpretation.

4.6. Examples of real-time grey-scale scanning

Figure 9 shows a large curvilinear transducer array being used to scan a pregnant woman andfigure 10 is typical of the kind of image that is produced.

Ultrasonic imaging has an important place in almost every area of clinical investigation.For more information, reference should be made to the numerous textbooks that are available:a good starting point is that edited by McGahan and Goldberg (1997).

4.7. Blood flow and tissue motion imaging

4.7.1. Ultrasonic scattering by blood.At the typical ultrasonic frequency of 3 MHz, thewavelength in blood is about 500µm. An individual red blood cell is a biconcave disc, witha diameter of about 8µm and a thickness of about 2µm. Scattering of ultrasound by bloodcan be modelled in several ways. For example, Brody and Meindl (1974) treated blood as asuspension of point scatterers. In fact, however, blood cells are quite closely packed and sothe individual cells do not behave like uncorrelated scatterers but actually interact strongly.This problem was avoided by Angelsen (1980), who modelled blood as a continuous mediumwith fluctuations in density and compressibility. Human blood cells have a tendency to formclumps, or ‘rouleaux’, which can survive even under normal flow conditions (Machiet al1983). Nevertheless, scattering tends to decrease with increasing shear rate (Yuan and Shung1989). With this as a model, Mo and Cobbold (1986) concluded that the backscattering ofblood can be considered to be a Gaussian random process.

If blood really did consist of a suspension of uncorrelated point scatterers and theultrasonic detection process was incoherent, it is arguable that blood flow could not bedetected by ultrasound. The ultrasonic power backscattered by the blood would remainconstant and there would be no discrete targets whose motion would either give rise toa Doppler shift frequency or whose displacement could be observed over time. In fact,however, scattering by blood is an example of the process that gives rise to speckle. Bloodbehaves as an array of ensembles that give rise to fluctuations in backscattered power that fade

Page 25: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 695

Figure 9. An ultrasonic scanner being used for an obstetrical investigation. The probe is the largecurvilinear array shown in figure 7.

sufficiently slowly to allow enough time for their motion to be observed (Atkinson and Berry1974).

4.7.2. The continuous wave Doppler method.Consider a beam of ultrasonic waves ofconstant frequency and amplitude travelling through the body and encountering a vesselcontaining flowing blood. The ultrasound detected as the result of backscattering fromstationary tissues has the same frequency as that of the transmitted ultrasound; thatbackscattered by the flowing blood has its frequency shifted by the Doppler effect (seesection 2.3). If the same transducer was used both to transmit the ultrasound (for whicha 10 V signal would be likely to be required) and to receive the backscattered ultrasoundfrom the flowing blood (producing a signal of about 10µV), the receiver would have toaccommodate a dynamic range of about 120 dB. This would be very difficult to achieve.Therefore, with continuous wave ultrasonic Doppler systems, it is usual for separate transducers(usually mounted side-by-side in the same probe) to be used for transmitting and receiving.Typically, the receiver then has only to accommodate a dynamic range of about 60 dB. A signalcorresponding to the Doppler shifted echoes is obtained by multiplying the transmitted andreceived signals, which can conveniently be done with a diode detector following the ultrasonicfrequency amplifier connected to the receiving transducer.

Page 26: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

696 P N T Wells

Figure 10. Scan of a fetus in late pregnancy, made using a large curvilinear array of the type shownin figures 7 and 9. In this longitudinal scan, the mother’s head is to the left. The uterine cavityoccupies most of the image: the fetal head is on the right, and the fetal body and one of the fetallegs and feet can be seen.

4.7.3. The pulsed Doppler method.The continuous wave Doppler method (see section 4.7.2)provides no explicit information about the distance between the ultrasonic transducer and themoving target. This information can be provided, however, by combining the pulse–echoprinciple (see section 4.1.1) with the Doppler method of motion detection (Wells 1969). Theprinciples are illustrated in figure 11. The master oscillator runs continuously at the frequencyof the ultrasound to be transmitted. The clock pulses are obtained by dividing down fromthe master oscillator frequency (to maintain phase coherence). These clock pulses triggerthe transmit sample-length monostable at the pulse repetition frequency of the system. Thismonostable generates a pulse that opens the transmit pulse gate for the period that ultrasoundneeds to be emitted by the transducer to produce the desired sample length. With an ultrasonicspeed of 1500 m s−1, this corresponds to 0.67µs mm−1. Echoes received by the transducer areamplified and fed to the phase quadrature detector (Nippaet al1975). This circuit separates thesignals according to whether they have frequencies higher (i.e., with leading phase) or lower(i.e., with lagging phase) than that of the transmitted ultrasound. Higher frequency signalscorrespond to flow towards the transducer, and vice versa for lower frequency signals. Theforward and reverse flow signals from the phase quadrature detector are fed to a heterodyneprocessor, where they are mixed with a signal at a pilot frequency. The pilot frequency ischosen to be greater than the highest reverse flow frequency signal. In the output from theheterodyne processor, the pilot frequency corresponds to zero flow velocity; reverse flowsignals have frequencies lower than the pilot frequency, whereas forward flow signals havehigher frequencies. This arrangement makes it simple subsequently to analyse the signals interms of their frequency spectrum. Following the transmission of each ultrasonic pulse, theoutput from the heterodyne processor consists of a wavetrain in which later time correspondsto greater depth. The sample depth monostable, which is also triggered by the clock pulse,introduces a time delay chosen by the operator to correspond to the echo delay time (about

Page 27: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 697

Figure 11. Block diagram of a pulsed Doppler system providing depth and length adjustment ofthe sample volume and phase quadrature detection of blood flow direction.

1.33µs mm−1, because the pulse has to travel to the target and the echo has to return) fromthe beginning of the sample volume from which it is desired to collect the Doppler signals.The receive sample length monostable then opens the receive pulse gate for a period equalto the transmitted pulse duration, so that the output from the receive pulse gate is actuallythe corresponding sample of what is, in effect, the Doppler signal. The amplitude of thissignal is held in the sample-and-hold circuit until it is updated by the sample derived from thenext transmitted ultrasonic pulse. The output from the sample-and-hold circuit is smoothed byfiltering, subjected to audio amplification and fed to the frequency spectrum analyser (typicallybased on the fast Fourier transform method) and displayed (e.g., on a strip chart).

4.7.4. Doppler flow and motion imaging.The purpose of methods of flow and motion imagingis to produce a two-dimensional (or three-dimensional) image of an anatomical structure, withthe presence of blood flow (or tissue motion) indicated in its correct spatial position by some

Page 28: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

698 P N T Wells

form of image coding (usually colour). The code can carry information about characteristicsof flow or motion (e.g., its speed, direction, velocity or quantity). The clinical value of themethod may be greatly enhanced if the process can proceed in real time or, at least, if it can befast enough to follow physiological motion or changes in the position of the scanning probe.

The simplest way to produce an image of blood flowing in a vessel system is to makeuse of the fact that Doppler-shifted blood flow signals are detected only when the ultrasonicbeam passes through flowing blood. A continuous wave Doppler probe (see section 4.7.2) ismounted on a scanning system that measures the position of the probe in what is essentiallya two-dimensional plane, with the ultrasonic beam at an angle to the plane. The probe isscanned slowly by hand, in contact with the patient’s skin under which lie the blood vessels tobe imaged. Only where Doppler signals are detected does the instrument cause registrationsto appear on the display (Reid and Spencer 1972). The same principle can be extended tothree-dimensional imaging by using a pulsed Doppler system with a multiplicity of receivingchannels gated over a range of depths, with the probe mounted on a scanner providing three-dimensional measurement (Fish 1981).

The first two-dimensional blood flow imager (Reid and Spencer 1972) produced simpleblack-and-white maps, with blood flow shown as being either present or not present. It wasCurry and White (1978) who first produced images coded in colour to show the velocity ofblood flow. Their equipment used a logic circuit to exclude reverse-flow signals and three filtersto separate forward-flow signals into normal, moderately increased and markedly increasedfrequency ranges. By this means, the image was coded in red, yellow and blue. This allowedthe extent of localized narrowing of arteries, with associated increase in blood flow velocity,to be assessed. In Fish’s (1981) three-dimensional scanner, the direction of blood flow wasindicated by colour coding, with blue for flow towards the transducer and red, for flow away.This logical choice of colour corresponds to the colours of the light from stars, as originallyhypothesized by Doppler (1843).

Although manually-scanned ultrasonic Doppler blood flow imaging opened up a new areaof clinical investigation, it never came into widespread use. To use the technique successfully,a great deal of skill was needed. Also, as the technique was being developed and assessed,ultrasonic duplex scanning was introduced by Barberet al (1974). A duplex scanner is onethat enables two-dimensional ultrasonic pulse–echo imaging to be used to guide the placementof an ultrasonic beam for Doppler signal acquisition, thus to identify the anatomical locationof the origin of the Doppler signal. Duplex scanners can be based on either mechanical(see section 4.1.3) or transducer array (see section 4.2) real-time scanning systems. For theacquisition of Doppler signals, the ultrasonic beam has effectively to be stationary and to dwellin the appropriate position sufficiently frequently and for a sufficiently long time during theacquisition period to allow adequate sampling of the target motion. The moving parts of amechanical scanner cannot be stopped and started rapidly and so only a stored image is usuallyavailable during Doppler signal acquisition. This can be a problem if there is movement ofthe patient or the probe. Scanning of the beam of an array scanner can be stopped and startedinstantaneously, so simultaneous operation is possible with this type of system.

The arrival of the duplex scanner had a considerable impact on the practice of ultrasonicimaging. It opened up important new areas of clinical investigation, particularly in the heart andthe vascular system. It had none of the inconvenience of the manually-scanned flow imagingtechniques and these largely fell into disuse. An example of a duplex real-time grey-scale scanand the simultaneously-acquired Doppler frequency spectrum is shown in figure 12.

The basic principles of combining pulse–echo and Doppler two-dimensional images areillustrated in figure 13. Essentially, the idea is an extension of duplex scanning but withmultiple Doppler receiving channels. Using a modified duplex scanner, Brandestini and

Page 29: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 699

Figure 12. Duplex scan of a carotid artery, one of the main arteries supplying blood to the head,which is on the left of the grey-scale image. The scan was made at the region in the neck where theartery bifurcates into the external carotid, which supplies the face, and the internal carotid, whichsupplies the brain. The oblique line represents the direction of the ultrasonic beam selected for theacquisition of the Doppler signals from the sample volume, indicated by the broad bar. The markerpassing through the sample volume was adjusted by the operator to align with the artery, to enablethe machine to estimate blood flow velocity from Doppler shift frequency: see equation (11). TheDoppler frequency spectrum is displayed on the right. The vertical scale indicates the blood flowvelocity (m s−1); the large markers on the horizontal scale indicate time in 1 s intervals. Note thatblood flow does not reverse direction in the carotid artery during diastole.

Forster (1978) obtained a real-time two-dimensional pulse–echo image through which theyswept the ultrasonic beam of a 128-point pulsed Doppler system to superimpose blood flowsignals, colour coded in red and blue, on the grey scale anatomical image. The system wasdeveloped until it was eventually possible to operate at four frames per second (Eyerset al1981). What ultimately limited the speed was the process of Doppler frequency estimationthat was employed.

It was the publication in English of a full paper (Kasaiet al 1985) that drew widespreadattention to the results of work that evidently had been in progress in Japan for several years. Itshowed that autocorrelation detection provided a rapid means of frequency estimation and madeit possible for colour flow imaging to be performed in real time. The basis of the autocorrelationdetector is that the echo wavetrains from stationary targets do not change with time, whereassequential echo wavetrains from moving targets have corresponding changes in relative phase.As shown in figure 14, the autocorrelation detector produces an output signal that depends onthe relative phases of consecutive pairs of received echo wavetrains. Thus, the echo wavetrainsthemselves are their own references for phase comparison. The autocorrelation detectorfunctions by multiplying two echo wavetrains, one currently being received by the transducerand the other, having been received from the immediately preceding pulse transmission anddelayed for a time exactly equal to the interval between pulse transmissions. The output fromthe autocorrelator has constant amplitude except where consecutive wavetrains have phase

Page 30: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

700 P N T Wells

Figure 13. Block diagram of a real-time two-dimensional colour flow imaging system. In thisexample, a phased array transducer is used to produce a sector scan. (Alternatively, a linear arraytransducer can be used to produce a scan in parallelogram format.) The image formatter andscan converter accepts the colour and grey-scale signals together with scan position data from theultrasonic beam steering circuit. The operation of the system is synchronized by a clock pulsegenerator (for simplicity, this is not shown).

Figure 14. Block diagram of an autocorrelation detector and associated gating, scan-convertingand colour-processing circuits. The multiplier performs the process of autocorrelation.

differences. In figure 14, the separate processes of velocity and velocity variance calculationare indicated; the value of the velocity variance can be considered to be a measure of the widthof the Doppler frequency spectrum, which increases with the degree of flow disturbance.

The colour processor in figure 14 assigns luminance, hue and saturation to the display,following one of the schemes described in section 4.7.6.

4.7.5. Flow and motion imaging by time-domain processing.Doppler flow and motionimaging, as described in section 4.7.3, is a narrow frequency band method in which processingis performed in the frequency domain. Embree and O’Brien (1985) and Bonnefous and Pesque(1986) independently described how flow and motion information can also be obtained frombroadband ultrasonic echoes by means of time-domain processing. The method depends ontemporal tracking of the spatial position of individual coherent blood ensembles or tissueconstituents. In principle, it can be applied directly to the amplified ultrasonic signals orto the video signals as they appear on the display, although the implementations of the twoapproaches are quite different.

Time-domain processing of the ultrasonic signals enables blood flow (or moving tissue)velocity to be determined by a one-dimensional correlation between the echo wavetrainsacquired with consecutively transmitted ultrasonic pulses. The timet required for an ultrasonicpulse to complete the round trip between the transducer and a moving scatterer situated a

Page 31: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 701

distancez from the transducer is given by

t = 2z/c. (18)

If the interval between pulses (which is equal to the reciprocal of the pulse repetitionfrequency) is equal toT andv is the speed of the scatterer, then

v = −τc/2T , (19)

whereτ is the change in the time of arrival of the echo from the scatterer between consecutiveultrasonic pulses. The time-domain formulation is clearly similar to the frequency-domain(Doppler) formulation given in equation (11) and would need to include the cosine function inthe same way if the scatterer was crossing the ultrasonic beam at an angle to its axis. The crosscorrelation (the function that is performed by the multiplier in the time-domain realization offigure 14) can be achieved by shifting the relative time positions of consecutive wavetrains;when the product has its maximum value, the two wavetrains are shifted by (T + τ ) and, sinceT is known,τ can be estimated. The other principal difference between frequency- and time-domain realizations of the circuit in figure 14 is that, for time-domain processing, the digitalsampling frequency needs to be increased by a factor of about ten to provide the necessarybroadband capability.

Doppler flow imaging (see section 4.7.4) came into widespread clinical use well before thefeasibility of time-domain processing became apparent. Time-domain processing is somewhatfaster, however, it has better spatial resolution and it is not subject to an artifact correspondingto the Nyquist limit (see section 4.7.3). Time-domain processing is usually less sensitivethan the Doppler method and the relative risks of the theoretical biological hazards of high-intensity short-pulse and low-intensity delivery of similar quantities of energy require furtherinvestigation (Wells 1995; and see section 5).

It is the time-domain processing of the video signals that is perhaps the most obviousmethod of determining the velocity of blood flow or tissue motion. If a target can be visualizedby real-time two-dimensional imaging (in which the image is actually formed by the videosignals), its velocity can be estimated directly from the measurement of the distance that ittravels over a known interval of time. For example, Traheyet al (1987) tracked blood flow insequential two-dimensional image frames by means of a localized two-dimensional specklecorrelation search.

Another approach, applicable to blood flow studies but not to solid tissue motion, makesuse of the decorrelation of the speckle pattern, which occurs at a rate that depends on the bloodflow velocity (Gardiner and Fox 1989). The potential of this method remains to be explored.

4.7.6. Colour coding schemes.Colour is characterized by luminosity (brightness or shade),hue (which is determined by the wavelength of the light, but which is, of course, a purelysubjective phenomenon for the observer), and saturation (which is the degree to which a colourdeparts from white and approaches a pure spectral colour). The wavelengths of the visiblespectrum extend from about 390 nm (violet) to 740 nm (red). Under favourable conditions,an observer with normal colour vision can discriminate between hues in the middle part of thespectrum with wavelength differences of around 1.2 nm; about 130 steps of hue difference canbe perceived across the entire visible spectrum. As the saturation of a hue is decreased to forma tint, about 20 different levels can be identified.

Two colour coding schemes are in common clinical use for blood flow studies. In ‘colourvelocity imaging’, the image is colour-coded according to the velocity of blood flow. Increasingflow velocity towards the probe is coded red-orange-yellow-white; increasing flow velocityin the opposite direction is displayed in dark blue-blue-light blue-white. The colour scales

Page 32: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

702 P N T Wells

Figure 15. Heart scan made with the probe positioned on the anterior thorax at the lower edge ofthe rib-cage. The left ventricle is the large dark area near the apex of the scan. During diastole,blood from the left atrium, which is the chamber below the left ventricle in this scan, enters the leftventricle through the mitral valve. In the normal, the mitral valve closes in systole and blood ispumped through the aortic valve into the aorta. This scan was made during systole, and it can beseen that the mitral valve is not fully closed: this allows blood to flow back from the left ventricleinto the left atrium, as is evident from the green colour which represents the Doppler-shifted signalfrom the reverse blood flow (see colour bar on the right). The severity of the condition, whichis known as ‘mitral regurgitation’, can be judged from the area of the colour-coded region. Thedistance markers are separated by intervals of 10 mm.

can be interpreted without the need to refer to a coding key. Perversely, however, the colourdirections are the opposite to those that Doppler (1843) hypothesized for the light from stars,and adopted by Fish (1981) in his colour flow scanner (see section 4.6.3). In addition tocolour coding according to velocity, colour can be used to indicate the variance of the velocityestimate, which is related to the degree of flow disturbance. This can be done, for example,by restricting the limits of the velocity image to the reds and blues and injecting an increasingamount of yellow as the flow disturbance increases. The yellow mixes with red to form orangeand with blue to form green.

Examples of the clinical applications of colour velocity imaging are shown infigures 15–17.

In ‘colour power imaging’, which is the other principal method of colour flow imagedisplay, red or blue luminosity (depending on the preference of the observer) is used toindicate the power, or amplitude, of the blood flow signal. By appropriately selecting thegain of the system, stationary echoes are displayed with low luminosity, whilst blood flowappears proportionately brighter as the signal power increases (Rubinet al1994). The methoddoes not distinguish between forward and reverse flow. Its principal advantage is that it ismore sensitive than colour velocity imaging. This is because, as the system gain is increasedto increase the sensitivity in colour velocity imaging, the entire spectrum of colour begins toobliterate the display, as the broadband noise level is approached. In colour power imaging,noise merely increases the background image luminosity; blood flow signals, even thoughonly weak, cause a proportionate further increase in the localized luminosity. An example of

Page 33: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 703

Figure 16. Colour-velocity image of the umbilical cord of a fetus, floating in the amniotic fluidwithin the uterine cavity. The cord contains two arteries and one vein, which can be distinguishedin the scan by the directional colour-coding (indicated in the colour bar on the right). The distancemarkers are separated by intervals of 5 mm.

Figure 17. The brain is supplied with blood by the internal branches of the two carotid arteries,one on each side of the neck. In this colour velocity scan, the patient’s head is on the left. Thecarotid artery is partially obstructed by atheromatous plaque, which results in flow disturbance:this is evident from the mosaic of colours, in contrast with the uniform red in the undiseased part ofthe artery. The blue colour represents blood flowing from the head, in the jugular vein. The colourbar shows the coding scheme and the distance markers are separated by intervals of 5 mm.

a colour power image is shown in figure 18.Scanners used for colour-coded imaging of blood flow have a control (typically labelled

Page 34: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

704 P N T Wells

Figure 18. Transverse scan of kidney, made with the probe on the patient’s back, a few centimetresto the left of the spine. The blood flow is colour-coded according to the power of the Doppler-shiftedsignal. Although the coding scheme is not affected by the direction of the blood flow (the blue inthe colour bar is not assigned), the renal artery and renal vein can be seen lying side-by-side in thelower right corner of the colour box. The small vessels in the outer cortex of the kidney can alsobe seen, which would probably not be possible if the colour-coding was according to the velocityof the blood flow. The distance scale is in centimetres.

‘colour write priority’) that allows the operator to inhibit the display of colour over those areasof the image that are characterized by the presence of higher amplitude echoes than those thatare obtained from blood. This prevents the appearance of colour due to the motion of solidtissues. Under some circumstances, however, information about the motion of solid tissuescan be of clinical value. To provide a colour-coded image of solid tissue motion, the colourwrite priority is adjusted to inhibit the display of colour from image areas with low amplitudeechoes: only high amplitude echoes are displayed, colour-coded according to their velocity orpower. The process is sometimes called ‘tissue Doppler imaging’ (McDickenet al 1992). Aswith colour flow imaging, colour tissue motion imaging may be coded according to velocityor power (Hoskins and McDicken 1997).

Colour Doppler imaging, whether of flow or tissue motion, is sensitive to the angle betweenthe direction of the ultrasonic beam and the target movement, as indicated in equation (11). Itis possible, however, to determine the direction of target movement by measuring its velocityfrom several different positions on the scan plane. Thus, it is possible to produce a two-dimensional image in which the colour represents the speed of target movement (Hoskins1997).

The distinction between blood flowing in a large vessel and in solid tissue with theminimum of capillary blood flow to maintain viability ignores the intermediate situation inwhich there is significant perfusion and a multiplicity of small vessels within solid tissue.For example, malignant tissue is characterized by numerous small vessels with direct shuntsbetween arteries and veins: the associated ultrasonic Doppler blood flow signals are well

Page 35: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 705

correlated with the presence of cancer (Wellset al1977). By carefully adjusting the upper andlower thresholds of Doppler signal amplitude between which colour is displayed in the image,the extent of tissue vascularity can be estimated by counting the proportion of coloured pixels(Cosgroveet al 1993).

4.8. Three-dimensional image acquisition and display

Baum and Greenwood (1961) prepared a series of photographically-reversed (i.e., the higherthe echo amplitude, the darker the registration) contiguous two-dimensional ultrasonic scansof the eye, on transparent plates. By stacking these plates with appropriate spacing, theyconstructed a three-dimensional image of the eye. Such an image enables even a relativelyunskilled observer to appreciate ultrasonic information in a three-dimensional form. Withthis method, however, the deeper echoes are visible only poorly because they are obscured bysuperficial echoes.

Although the pioneering work was followed by other attempts at ultrasonic three-dimensional image acquisition and display, with a number of small improvements such asholography (Redmanet al 1969) and stereoscopy (McDickenet al 1972), real progress wasnot possible until digital image storage and manipulation became practicable (Halliwellet al1989).

Imaging involves the three stages of acquisition, processing and display. For a typicalpenetration of 150 mm, the pulse repetition rate might be 5000 s−1. As explained insection 4.1.3, this corresponds to an image frame rate of 20 s−1 with 250 lines per frame.The important point is that two-dimensional ultrasonic scanning can be in real time, with thedepths of penetration and image line densities that are needed to be clinically useful. It is truethat the image frame rate has to be reduced in duplex and, to a greater extent, in colour flowscanning (see section 4.7) but, even so, one of the important advantages of two-dimensionalultrasonic imaging is that it is essentially a real-time process.

Three-dimensional ultrasonic imaging requires the acquisition of a set of lines of imageinformation (echo wavetrains) not just from a two-dimensional tissue plane, but from a three-dimensional tissue volume. Imagine that a volume of tissue 100 mm2 × 150 mm deep isto be scanned in three dimensions, by acquiring a set of contiguous two-dimensional scans.Allowing for the fact that the spatial resolution in elevation is likely to be around three timesworse than that in azimuth (see section 4.4), 33 contiguous scan planes each of 100 lineswould be appropriate. This means that 3300 lines would need to be acquired and, at a pulserepetition rate of 5000 s−1, 1.5 image data sets could be acquired per second. Although thisis not fast enough to qualify as real time, it certainly is adequate to be clinically useful forthe investigation of relatively static anatomical structures. Also, by synchronizing the imageacquisition with the electrocardiogram, satisfactory three-dimensional scans of the heart in thevarious phases of the cardiac cycle can be acquired in a few seconds. Contemporary techniquesin x-ray computed tomography, magnetic resonance imaging and nuclear medicine cannot evenremotely approach the speed of three-dimensional ultrasonic imaging.

Figure 19 shows three ways in which three-dimensional image data can be acquired with aphased array sector scanner. All these methods can be implemented by a motorized movementof the transducer within a hand-held device. Another approach is by free-hand scanning witha transducer designed for two-dimensional imaging, with the position and orientation of theprobe being measured in three-dimensional space. Although the operator needs to be skilledin order to acquire a regularly-spaced image data set, the probe has hardly to be any morebulky than for traditional scanning. It is prudent to ensure that the spatial resolution obtainablein two-dimensional scanning is maintained in the three-dimensional image and, in abdominal

Page 36: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

706 P N T Wells

Figure 19. Three-dimensional image data acquisition with a phased array scanner. (a) Rotation ofthe transducer about an axis normal to the centre of the array scans a conical volume of tissue. (b)Rotation of the transducer about the longitudinal axis of the array scans a pyramid-shaped volumeof tissue. (c) Translation of the array along an axis normal to the longitudinal axis of the arrayscans a wedge-shaped volume of tissue.

imaging, this means that the spatial measuring system typically needs to have positional andangular accuracies of better than around 0.5 mm and 0.5◦, respectively. Articulated arms withangle encoders, acoustic methods using spark gaps and microphones, and light-emitting diodeswith stereoscopic TV monitoring, have all been tried. Currently, however, the most popularmethod seems to be that based on electromagnetic field transmitters and receivers (Detmeret al 1994, Barryet al 1997). In this method, the receiving sensor is typically a 10 mm cube,containing three orthogonal coils. It is attached to the ultrasonic probe. The transmitting coilsare mounted in a fixed spatial relationship with the couch on which the patient lies. Thesecoils are pulsed in sequence to emit three orthogonal magnetic fields. The process is repeatedfast enough to track even rapid changes in the position and orientation of the ultrasonic probe.Typically, a skilled operator can acquire a good grey-scale image data set in around 5 s, or acolour flow image data set in around 20 s.

In principle, it is possible to acquire a three-dimensional image data set without anymechanical scanning, by the use of a two-dimensional transducer array. This is because suchan array can steer the ultrasonic beam both in azimuth and in elevation. Progress has beenmade in the construction and operation of such arrays (Smithet al1991, von Rammet al1991)but formidable problems remain, mainly due to the large number of small transducer elementsthat are required.

An image data set, whether acquired by automatic or by manual scanning, can beconsidered to be a three-dimensional block of small volume elements (called ‘voxels’), thebrightness or colour of which carry the same information as the picture elements (‘pixels’) inan ordinary two-dimensional scan. An important difference from two-dimensional imaging,however, is that the complete three-dimensional data set cannot be collected in true real timewith contemporary technology.

A three-dimensional ultrasonic image data set can be processed and displayed on acomputerized workstation of the kind currently used in radiology for three-dimensional x-ray computed tomography and magnetic resonance imaging (Fishmanet al 1991). Anexample of such a display is shown in figure 20. Unlike three-dimensional CT and MRI(in which, for example, bone can easily be distinguished from soft tissue), however, ultrasonic

Page 37: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 707

Figure 20. Display of three-dimensional scan of ovary obtained with an endovaginal probeincorporating a curvilinear transducer array, mechanically rotated through a sector in the planeorthogonal to that of the array scan. The three-dimensional image can be displayed in sections withpositions and orientations selected by the operator, either in real time or from the stored data.

images generally cannot satisfactorily be automatically ‘segmented’ into separate anatomical orstructural elements. (An exception is the lumens of blood vessels, where these are characterizedby the presence of colour-coded flowing blood.) This is because ultrasonic images areessentially ‘noisy’, consisting largely of speckle due to the coherent nature of ultrasound (seesection 4.5). Consequently, surface fitting does not work well with grey-scale images (althoughprogress is being made in developing active contour models (Chalanaet al 1996), integratededge maps (Aarniket al1998) and minimum cross-entropy thresholding (Zimmeret al1996))and segmentation methods are the subject of much contemporary research. Three-dimensionalultrasonic angiograms, with power Doppler as the segmentation tool, are already being usedclinically (Richieet al1996); they indicate what may become possible with grey-scale imagesin the future.

Following successful segmentation, the ‘surface rendering’ display method can be used.Lines-of-sight, or rays, are constructed from any chosen observation point and a perspectiveview of the intersections of the segmented surface by these rays is displayed (Carsonet al1992). Stereoscopic viewing or rotation of the image gives the illusion of three dimensions.

Even when satisfactory segmentation is not possible, ‘volume rendering’ may be used todisplay three-dimensional ultrasonic images (Nelson and Pretorius 1998). As with surfacerendering display, rays are cast from a point. The brightness (or colour) of each ray, however,is determined by the local image brightness (or colour) as the ray passes through the imagevolume. The image volume can then be viewed from any chosen direction and from anychosen distance. In this way, stereoscopic pairs of images can be created, or the viewpointof the two-dimensional display of the observed image volume can be rotated to create depthcues. Unless deep structures are entirely surrounded by a brighter shell, it is likely that theyall will become visible at least during part of the cycle of rotation.

Page 38: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

708 P N T Wells

Two-dimensional imaging of the velocity of tissue motion or blood flow using time-domainprocessing is discussed in section 4.7.5. The time-domain processing method based on echotracking has been extended to three dimensions by Bohset al (1995), who have presentedimages of blood flow characteristics in laminar conditions and in jets.

From three-dimensional imaging, it is a simple step, in principle, to four-dimensional(space and time) imaging. Preliminary studies of four-dimensional imaging of the fetal hearthave been carried out (Denget al1996). The data were acquired over about 20 s, using a gatingsignal derived from a fixed-position pulse–echo ultrasonic beam for stroboscopic visualization.

4.9. Specialized imaging methods

4.9.1. Endoluminal scanning. ‘Endoluminal scanning’ is the term used to describe anymethod of imaging in which scanning is accomplished by means of a probe positioned withinan anatomical cavity. Such a cavity may have a natural opening (e.g., the oesophagus, therectum and the vagina) or access may require tissue puncture (e.g., the arteries, the veins, theabdominal cavity and the chambers of the heart).

The principal advantage of endoluminal scanning is that it brings the transducer into closerproximity to the structures to be scanned than would otherwise be possible. This means thathigher ultrasonic frequencies can be used and this, together with the smaller thickness ofintervening tissue to distort the ultrasonic beam, results in images with better resolution andclarity. The main disadvantage is that endoluminal scanning is a more invasive procedure.

Probes for endorectal and endovaginal scanning are usually based on small (10–20 mm×5 mm) high frequency (5–10 MHz) linear or curved linear transducer arrays (seesection 4.2) mounted at the end of a suitably-shaped rod, the other end of which is fitted with ahandle for the operator. Because these probes are designed to scan within cavities with naturalopenings, they do not need to be sterilized before use: it is sufficient for them to be cleanedwith antiseptic solution. An example of an endovaginal scan of an early pregnancy is shownin figure 21.

For endo-oesophageal scanning, a small transducer is mounted at the tip of a long (upto 1 m) flexible tube, with a maximum diameter of about 10 mm. The endoscope maycarry an optical imaging channel and there may also be provision for surgical instrumentsto perform interventions under direct vision. Radial scanning around the axis of the endoscopeis accomplished by mechanical rotation of the transducer, which typically has a diameter of3–5 mm, has lens focusing of the ultrasonic beam and operates at 5–10 MHz. The patient,who is usually sedated, has to swallow the tip of the endoscope; a little local anaesthetic at theback of the throat minimizes the discomfort of the procedure. Since the endoscope does notpenetrate the skin, it needs only to be cleaned with antiseptic solution.

For intra-abdominal scanning, a procedure which is becoming increasingly common as anadjunct to minimally invasive laparoscopic surgery, rigid or flexible probes can be used. Forexample, a typical ultrasonic laparoscope has a 5–10 MHz linear transducer array, 30 mm by3 mm, mounted at the tip of a 50 cm rigid rod. The angle of the tip can be controlled in twoplanes by the operator, using knobs on the handle of the laparoscope, to select the scan plane.Since the laparoscope is introduced through a port through the abdominal wall, it has to besterilized before use. This is conveniently done by exposing it to an atmosphere of ethyleneoxide gas for 3–7 h, followed by 2 h aeration for detoxification.

For intravascular scanning, there are two distinct approaches. Both employ a flexiblecatheter about 1.2 m long and with a diameter of about 3 mm, with a 20–30 MHz transducermounted at the tip to produce a radial scan: in one approach, the beam of a single-elementtransducer is mechanically rotated (Wells 1966) and, in the other, the beam of a cylindrical

Page 39: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 709

Figure 21. Scan of early pregnancy made using an endovaginal probe such as that shown in figure 7.The uterine cavity is the dark area in the middle of the picture. The fetus lies close to the probe.The length of the fetus (crown-to-rump) extends between the two markers, which were positionedby the operator. The scanner incorporates a look-up table that displays the corresponding timesince conception (in this case, 6.5 weeks). The distance scale is in centimetres.

array of (typically 64) transducer elements is electronically rotated (Bomet al1972, O’Donnellet al 1997). Mechanical scanning is usually accomplished by rotating the transducer througha flexible drive-shaft by means of a motor mounted at the other end of the catheter, as shownin figure 22 (Wells 1966, ten Hoffet al 1995), although micromotors small enough to beincorporated into the catheter have recently been described (Erbelet al 1997). The catheteris introduced into an appropriate blood vessel through a skin puncture and advanced until thetip is positioned to scan the region of interest. For example, a femoral artery puncture in theleg affords access to the coronary arteries in the heart. Examples of such scans are shown infigure 23.

Endoluminal scanning is usually used for two-dimensional imaging. The operator canobtain a three-dimensional impression of the scanned anatomy by changing the scan plane.The scan plane can also be adjusted automatically in known increments, e.g., by pulling-backan intravascular catheter, to acquire an image data set for three-dimensional display (Liet al1995).

4.9.2. Synthetic aperture imaging.Synthetic aperture imaging depends on the synthesis ofthe equivalent of a large aperture transducer by means of a small transducer which is scannedto fill the aperture whilst acquiring pulse–echo information from the tissues to be imaged. It isa two-stage process. In the first stage, the tissues are scanned from a line or an area, by a line(fan beam) or point (conical beam) transmitter, and the scattered ultrasound is itself scannedby a line or a point receiver. In a monostatic synthetic aperture, the same transducer is usedas the transmitter and the receiver. The technique has been tried for ultrasonic body scanning

Page 40: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

710 P N T Wells

Figure 22. Intravascular catheter probe for scanning blood vessels from within their lumens. Leftpanel: detail of the scanning tip, showing the transducer, in side view, set a slight angle to themechanical drive shaft axis, to reduce the amplitude of the echo from the walls of the catheter;the catheter is filled with liquid to provide ultrasonic coupling. Right panel: the complete deviceshowing the length of the catheter, the scanning tip and the connector (with slip rings, for electricalconnexion) for the motor drive to rotate the transducer.

(Burckhardtet al 1974, Ylitalo and Ermert 1994). Because a diverging beam has to be used,the sensitivity decreases rapidly with depth. The sensitivity increases as the divergence isreduced, but this also results in reduced spatial resolution (Ylitalo 1996). Nevertheless, theimaging performance is theoretically better than that of a traditional B-scanner, although thetime required to synthesize the aperture may mean that physiological motion degrades theimage. Moreover, tissue inhomogeneity reduces the coherence of the ultrasound, on which themethod depends.

Synthetic aperture processing does have a useful role in intravascular ultrasonic scanningwith a small cylindrical transducer array (see section 4.9.1) and it may also reduce the imageacquisition time in three-dimensional ultrasonic microscanning (see section 4.9.5).

4.9.3. Computed tomography.The invention of x-ray computed tomography (Hounsfield1973) had a profound effect on the practice of clinical radiology. The method depends onthe acquisition of an angular set of x-ray attenuation profiles from a two-dimensional tissueplane, from which the image is produced by the process of back-projection reconstruction(Ramachandran and Lakshminarayanan 1971). It works well partly because x-rays travel instraight lines. In principle, ultrasonic computed tomography depends on the acquisition of acomplete set of projections of the characteristic to be imaged (i.e., velocity or attenuation).

Figure 24 shows how ultrasonic computed tomography can be used for imaging thebreast. Ultrasonic computed tomography depends on the validity of the assumption thatline-of-sight propagation is maintained across the tissues being scanned. Unfortunately,however, an ultrasonic beam is deviated by refraction and distorted by inhomogeneities intissue. Consequently, and beginning with the work of Greenleafet al (1974), who scanned anexcised dog heart at 5 MHz, the results have until recently been consistently disappointing. CT-

Page 41: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 711

Figure 23. Intravascular ultrasonic scans of coronary arteries. The scans were made with a high-frequency cylindrical array transducer with a diameter of about 3 mm, introduced via a punctureinto the femoral artery in the leg. The circle in the blood vessel lumen represents the transducerarray. Top panel: normal artery. Bottom panel: artery affected by eccentric atheromatous plaque(indicated by arrows).

derived data on speed-of-ultrasound, however, has been used by Jago and Whittingham (1991)to correct misregistration in superimposed B-scans obtained from multiple angles around anexcised sheep kidney, with encouraging results.

4.9.4. Elasticity imaging. Doctors all learn the art of paplation as an essential element ofclinical examination. Palpation depends on the differences in the hardness of different tissuesand thus can reveal the presence of abnormalities if they are close to the surface. An obviousexample is the discovery of a cancer of the breast, detectable because malignant tumour tissueis harder than normal glandular and fatty tissues.

The goal of ultrasonic elasticity imaging is to map tissue properties such as Young’s

Page 42: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

712 P N T Wells

Figure 24. Ultrasonic computed tomography. In this example, time-of-flight tomography ofa breast immersed in water is being performed by translate-rotate scanning. (a) Mechanicalscanning system. (b) Profiles of time-of-flight measurements obtained by translation at threeangular positions: in a real system, such profiles are acquired at, typically, 128 angular positionsaround the entire tomographic plane.

modulus (or stiffness), Poisson’s ratio and viscosity in an anatomically meaningful presentationto provide useful clinical information (Gaoet al1996). The stress–strain relationship for mosttissues is nonlinear and stress tends to relax over time under constant strain (i.e., cyclic loadingand unloading is characterized by hysteresis).

Several methods of ultrasonic elasticity imaging have been demonstrated. In vibrationamplitude sonoelastography, a low-frequency (20–1000 Hz) vibration is externally applied toexcite internal tissue motion of which an image is produced by Doppler detection (Lerneret al 1988). In vibration phase gradient sonoelastography, both the amplitude and the phaseof externally-excited low-frequency internal tissue motion are measured. By assuming thatviscosity at low frequencies is negligible and that shear waves predominate, Levinsonet al(1995) obtained phase gradient images of thigh muscle under various conditions of activemuscle contraction: both the speed of vibration propagation and the value of Young’s modulusincrease with increasing contraction. The method of compression strain elastography, inwhich the tissue is externally compressed and pre- and post-compression ultrasonic A-scanline pairs are crosscorrelated to produce a set of strain profiles, has been demonstrated byOphir et al (1991) and Kallelet al (1998). In an ingenious variant of the method, de Korteet al (1998) used an intravascular imaging system to obtain strain elastograms of diseasedarteriesin vitro, with the change in strain being produced by change in pressure. With furtherrefinement, this method may be practicablein vivo, by making use of the arterial pressurepulse. O’Donnellet al (1994) have devised a method of compression strain elastography withrelatively large displacements (around ten wavelengths), measuring the overall displacementby summing the small displacements resulting from incremental step loading from the A-scan crosscorrelations. Finally, Bohs and Trahey (1991) have used a two-dimensional speckletracking method employing a sum-of-absolute-difference criterion with a search kernel tomeasure flow and tissue motion and this has been adapted by Walkeret al (1993) for vibrationamplitude sonoelastography.

Ultrasonic elasticity imaging has not yet been used, except rather crudely, in routine clinicalpractice. It is a very promising method, however, because it should have spatial resolution at

Page 43: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 713

Figure 25. Geometry of the ultrasonic beam produced by aspherical bowl transducer.

least comparable with that of real-time grey-scale imaging combined with potentially bettertissue discrimination.

4.9.5. Microscanning. Microscanning is defined as the two- and three-dimensional displayof biomedical soft tissue structures with spatial resolution in the range 10–100µm. Thisresolution range lies between the smallest structures that can be seen under direct visionand the largest structures that are traditionally of interest to histopathologists using the lightmicroscope. Although resolutions exceeding that of the light microscope can be obtained withthe scanning acoustic microscope (Lemons and Quate 1974) operating at frequencies above3000 MHz (i.e., with wavelength of less than 500 nm), the need to section the tissue, the longscanning times and the relatively high cost may be the reasons why this has not been introducedinto clinical practice. The scanning laser acoustic microscope (Kessler 1974), which is a real-time instrument in which the image is acquired by laser-scanning a liquid surface levitated byultrasound transmitted through the thin specimen, operates at frequencies of around 100 MHz(i.e., with wavelengths of around 15µm). The need for sectioning and the high cost haveprevented the method from becoming popular. Microscanning, with its ability to exploretissues in three dimensions without the need for serial sectioning, has previously been largelyneglected but has the potential to advance biomedical science and clinical practice.

The geometry of a focused ultrasonic beam produced by a spherical bowl transducer isshown in figure 25. The dimensions of the−3 dB profile of the ellipsoidal focal volume canbe calculated from the following equations (Wells 1977):

df ≈ λ(F/2a) (20)

lf ≈ 15(1− 0.01θ)df , (21)

wheredf is the cross-sectional diameter andlf is the axial length of the focal volume,a is theradius of the spherical bowl andF is the focal length of the transducer, andθ is the half-angleof convergence of the beam, provided thatθ < 50◦. The focal beam diameter can be used asthe starting-point for the design of an optimized microscanner. This is because it determinesthe resolutions in elevation and in azimuth; the range resolution is not determined by the lengthof the focal volume, but by the length of the ultrasonic pulse and, generally, this can be madeless than the beam diameter.

Page 44: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

714 P N T Wells

A typical scanning acoustic microscope operating at a frequency of 100 MHz, using atransducer with diameter of 3.2 mm and a focal length of 6.4 mm, thus hasdf ≈ 30µm andlf ≈ 380µm. This means that its optimal resolution cell is around 30µm in diameter and, bygating the echo wavetrain into separate receiver channels, 13 appropriately-spaced resolutioncells can be accommodated axially within the focal volume. For imaging a 5 mmtissuecube from opposite sides in two 2.5 mm-thick slabs, sampling theory requires the scanningbeam spacing in theXY plane is equal to 15µm; in theZ direction, the process has to berepeated over about 13 sets of 13 planes, from each side. In order fully to sample the imageinformation from the 5 mm tissue cube, this means that the beam has to dwell to acquire echoeson about 2.9 million occasions. Assuming a dwell time of 1 m s−1 (chosen to allow achievablespatial scanning velocity and the possibility of signal averaging for noise reduction), the totalscanning time is about 48 min. At 100 MHz, the attenuation of ultrasound is about 20 dB cm−1

in water and about 100 dB cm−1 in soft tissue. Therefore, the overall dynamic range of gaincompensation needs to be about 40 dB for a 2.5 mm-thick slab of tissue.

It will be interesting to see whether three-dimensional miroscanning with these constraintsturns out to be clinically useful. If it does, it should be possible substantially to reduce theimage acquisition time by employing synthetic aperture scanning (see section 4.9.2).

4.9.6. Contrast agents. It was Gramiaket al (1969) who first reported that the ultrasonicechogenicity of blood could be artificially enhanced. They observed that the injection ofindocyanine green dye into the chambers of the heart by means of a catheter (a procedure used toopacify the chambers and vessels in x-ray imaging) coincidentally caused the blood transientlyto give rise to strong ultrasonic echoes. Subsequently, Kremkauet al (1970) showed that theeffect was not primarily due to the echogenicity of the indocyanine green, but to decompressioncavitation at the tip of the catheter. Two years later, Ziskinet al (1972) demonstrated that etherwas the most effective substance that they tested to enhance echogenicity, presumably becauseit boils vigorously at body temperature. In any but tiny quantities, however, ether is toxic oreven lethal. Water saturated with carbon dioxide was found to be satisfactory, since it containsnumerous bubbles but gas emboli are not a complication as the bubbles rapidly dissolve inblood.

Because gas has a very low characteristic impedance in comparison with that of blood(see table 1), suspended gas bubbles greatly increase the echogenicity of blood. The scatteringis amplified if the size of the bubble is such that it resonates at the frequency of the incidentultrasound. The resonant frequencyf0 of a gas bubble immersed in a liquid of densityρ andat pressureP is given by (Minnaert 1933)

f0 ≈ 12πr(3γP/ρ)

1/2, (22)

wherer is the radius of the bubble andγ is the adiabatic ideal gas constant. This means thata bubble with a diameter of 5µm (comparable with that of a red blood cell) resonates at afrequency between 1 and 10 MHz. The actual frequency depends markedly on the nature ofthe shell that encapsulates the bubble. Commercially-available microbubble contrast agentsare encapsulated in a variety of substances, such as albumin, lipid and palmitic acid. The shellsstabilize the bubbles and extend their life after injection into blood.

It is a fortunate coincidence that the bubbles that resonate at the ultrasonic frequenciescommonly used for imaging have diameters of a few micrometres. From the point of view ofclinical convenience and the safety of the patient, an injection into an artery is considered to bevery much less desirable than an injection into a vein. Blood enters the right side of the heartfrom the veins and then passes through the lungs before returning to the left side of the heart,which pumps it through the systemic circulation. Consequently, an intravenous injection of

Page 45: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 715

ultrasonic contrast agent soon enters the systemic arterial system, provided that the individualmicrobubbles are small enough to cross over the capillaries in the lungs. Obviously, red bloodcells do this; provided that the microbubbles are no bigger than red blood cells, they also cando so. The only condition is that they should survive for long enough to produce the desiredimage contrast.

Encapsulation of a microbubble results in an increase in its resonant frequency and adecrease in the amplitude of scattered ultrasound (de Jong and Hoft 1993). The lifetime ofthe microbubble can be extended either by increasing the stiffness of its shell or by usinga gas that dissolves poorly in blood (Frinking and de Jong 1998). Free air bubbles with adiameter of about 8µm disappear within 1 s; those with a thin albumin shell persist for upto 10 min; and those with stiff shells or relatively insoluble gases can last for an hour ormore.

The obvious utility of an ultrasonic contrast agent arises because it increases theechogenicity of blood and thus, for example, increases the sensitivity of ultrasonic Dopplerblood flow detection. It also results in a significant reduction in image clutter. Theconcentration of contrast agent needed to produce useful enhancement of blood echogenicityis typically 50 mg of concentrated bubble suspension per kg body weight. Used in thisway, microbubbles actually decrease the contrast between blood and the soft tissues thatsurround the blood vessel and so the term ‘contrast agent’ is something of a misnomer. Itis possible, however, to use microbubble contrast agents in a way that does increase theechogenicity of blood beyond that of soft tissues. One way might be to inject more contrastagent, but this would not be attractive because of the risk associated with introducing gasinto circulating blood, where it might block small vessels and, for example, cause a stroke.Another problem would be the increase in the ultrasonic attenuation in the blood (Bouakazet al 1998). The method that does work, however, exploits the fact that microbubblessuspended in blood backscatter ultrasound at harmonics of the incident frequency, whenthe incident pressure is sufficiently high (Schropeet al 1992). Consequently, if thereceiver is tuned to receive signals at twice the frequency of the transmitted ultrasound,the echoes backscattered by the contrast agent microbubbles in the blood can actuallyhave a higher amplitude than those backscattered by the surrounding soft tissues. Thisimproves the discrimination between blood and tissue echoes and is known as ‘harmonicimaging’ (Forsberget al 1996). It has been shown by Zheng and Newhouse (1998) that,for a given incident ultrasonic frequency, the backscattered signal commences with thesame frequency, but this is followed after an ‘onset delay’ of typically 15 cycles of thefundamental by the development of the second harmonic. This effect must limit the rangeresolution obtainable with harmonic imaging, although this does not yet seem to have beeninvestigated.

At first sight, it is rather surprising that a stronger backscattered signal is detected atthe second harmonic of the bubble resonant frequency, than at the third harmonic. Thisphenomenon has not yet been thoroughly investigated but a possible explanation is thatincreasing attenuation in overlying tissues (which increases with the frequency) and the limitedfrequency bandwidth of practical transducers are dominant. Another possibility is that the firstsubharmonic frequency may be strongly backscattered and Shankaret al (1998) have shownthat it should be more easily detectable than the second harmonic. Improved detectability hasto be weighed against reduced resolution, however, and this is also a topic for research in thefuture.

It has been reported (Porteret al 1997) that encapsulated microbubbles can be destroyedby the ultrasonic exposure conditions used in some diagnostic procedures. The phenomenonis accompanied by the emission of strong ultrasonic transients that can result in the appearance

Page 46: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

716 P N T Wells

of high brightness pseudoechoes in pulse–echo grey-scale images and multicoloured transientsin flow and motion images. Thus, it provides a technique further to increase the sensitivityof ultrasonic imaging systems to the presence of blood containing ultrasonic contrast agent.Furthermore, the annihilation of contrast agent by exposure to a brief pulse of ultrasound canprovide a relatively precise timing marker for the measurement of blood flow and perfusiondynamics.

In addition to the use of contrast agents for improved imaging by echo enhancement,contrast agents can also be used for dynamic studies. Thus, a rapid increase in image brightnessmay occur after injection of a bolus of contrast agent, followed by a slower decline (Sehgaland Arger 1997). Analysis of the brightness-time curve can provide data on blood flow rateand blood perfusion.

According to equation (22), the resonant frequency of a bubble is proportionalto the square root of its ambient pressure. Fairbank and Scully (1977) proposedthat a noninvasive method for intracardiac blood pressure measurement could be basedon this relationship. Such a method would be of very great clinical value, sinceblood pressure can otherwise be measured only indirectly or invasively. Unfortunately,however, there are several reasons that make the resonant bubble approach impracticable.Firstly, bubbles cannot be manufactured with sufficient accuracy to ensure that theyare all the same. Secondly, and of fundamental importance, the bubble resonanceis not sharp enough to provide the pressure discrimination necessary to be clinicallyuseful.

4.9.7. Tissue harmonic imaging.The reduction in image clutter that can be obtained byrestricting signal detection to the second harmonic frequency backscattered by ultrasonicmicrobubble contrast agents is discussed in section 4.9.6. The method is possible becausebubbles behave in a nonlinear fashion when the ultrasonic pressure amplitude is sufficientlyhigh. It is not only bubbles that demonstrate nonlinearity, however, and table 1 shows thatbiological soft tissues also have this property. Consequently, ultrasound backscattered bytissues has harmonic frequency components that can be detected if the incident ultrasonicpressure is sufficiently high. As with microbubble contrast agents, it is the secondharmonic frequency that is most relevant. There the direct similarity ends. In tissueharmonic imaging, the harmonics are not generated during the scattering process, butduring the passage of the ultrasonic pulse through the tissue towards the scatterer (andnot following scattering, because the ultrasonic pressure amplitude is then too smallfor nonlinearity to have any material effect). Thus, the process is that described insection 2.8.

By tuning the receiver to the second harmonic of the transmitted ultrasonic frequency,two advantages can be gained. The first is that echoes detected from tissues close to thetransducer are relatively weak, because second harmonic generation builds up to a significantlevel only as the result of adequate tissue penetration. The second is that contrast resolutionis improved (because the side lobes of the transmitting and receiving beams have differentangular structures) and the spatial resolution is improved (because the distortion of thetransmitted beam by the inhomogeneity of the superficial layers is relatively small and thereceiving beam can be made narrower than the transmitted beam). The practical realizationof tissue harmonic imaging is relatively new, but see, for example, Thomas and Rubin (1998)for a discussion of the method in the context of echocardiography. Figure 26 illustratesthe improvement in image quality that can result when tissue harmonic imaging is used inabdominal scanning.

Page 47: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 717

Figure 26. Longitudinal scans of the right kidney, which contains a fluid-filled mass, made withthe probe on the anterior abdominal wall immediately below the rib cage. The wedge-shapedstructure near the top of each image (i.e., close to the probe) is the liver. Right panel: traditionalsignal processing, with the transmitter and the receiver operating at the same frequency. Leftpanel: second harmonic imaging, with the receiving transducer operating at twice the transmittedfrequency. The distance markers are separated by intervals of 10 mm.

5. Safety considerations

Because bioeffects, some of which are harmful, may be caused by ultrasound under certainexposure conditions, there is a hypothetical possibility that ultrasonic imaging may not becompletely safe (Wells 1986). Moreover, the ultrasonic exposure levels used by commercially-available scanners have been steadily increasing, in order to obtain more information (Duckand Martin 1991). Consequently, both regulatory authorities and prudent clinicians take thesubject seriously.

The World Federation for Ultrasound in Medicine and Biology (WFUMB 1992) haspublished the following statements on thermal effects in clinical applications.

B-mode imaging: known diagnostic ultrasound equipment as used today for simple B-mode imaging operates at acoustic outputs that are not capable of producing harmfultemperature rises. Its use in medicine is therefore not contraindicated on thermalgrounds. This includes endoscopic, transvaginal and transcutaneous applications.

Doppler: it has been demonstrated in experiments with unperfused tissue that someDoppler diagnostic equipment has the potential to produce biologically significanttemperature rises, specifically at bone-soft tissue interfaces. The effects of elevatedtemperatures may be minimized by keeping the time for which the beam passesthrough any one point in tissue as short as possible. Where output power can becontrolled, the lowest available power level consistent with obtaining the desireddiagnostic information should be used.

Although the data on humans are sparse, it is clear from animal studies that exposuresresulting in temperatures less than 38.5 ◦C can be used without reservation on thermalgrounds. This includes obstetric applications.

Page 48: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

718 P N T Wells

Transducer heating: a substantial source of heating may be the transducer itself.Tissue heating from this source is localized to the volume in contact with thetransducer.

The possibility that nonthermal effects of ultrasound may be hazardous in some situationsis more contentious (Barnettet al 1994). Cavitation, defined as the formation or activityof gas- or vapour-filled cavities (bubbles) in a medium exposed to an ultrasonic field, is thephenomenon of most concern. Other possible nonthermal mechanisms include radiation force,acoustic torque and acoustic streaming.

Many questions relating to the safety of the ultrasonic exposures used for imaging remainto be answered. For example, is there a linear relationship between the quantity of imageinformation and the ultrasonic energy needed to obtain it? The thermal effect of a givenultrasonic power may be independent of the exposure duty cycle, but what are the nonthermaleffects under different regimes? Pragmatically, users of ultrasound for diagnosis should applythe ALARA (‘as low as reasonably achievable’) principle to the exposures to which they subjecttheir patients. The exposure levels should be at the lowest intensities and for the briefest timesnecessary to obtain diagnostically-adequate images. To assist in achieving this goal, a predictorof cavitation known as the mechanical index (MI) has been developed (AIUM/NEMA 1992),given by the expression

MI = Pr · 3(zsp)f1/2c , (23)

wherePr · 3(zsp) is the peak rarefactional pressure (MPa) derated by 0.3 dB cm−1 MHz−1 tothe point on the beam axis (zsp) where the pulse intensity integral is maximum, andfc is thecentre frequency (MHz). Some modern scanners display the MI value on the screen, so theoperator can aim to minimize it.

Although it is right to be concerned about exposure conditions, it is the consequencesof misdiagnosis that are likely to be the greatest hazard of an ultrasonic investigation (Wells1986). Other risks must not be ignored, but they must be viewed in the proper perspective.

6. Conclusions and future prospects

The physics of ultrasonic imaging is quite well understood. The performance of pulse–echo grey-scale imaging systems is ultimately limited by the attenuation, nonlinearity andinhomogeneity of tissues and by the need to minimize the possibility of hazard by minimizingthe ultrasonic exposure. Despite these limitations, however, there is scope for improvement.The rate of improvement in the capabilities of digital electronic circuits shows no sign ofdiminishing and, as costs continue to fall, so ultrasonic signal digitization is moving closer to theindividual transducer element. The potential benefits of contrast agents are only just beginningto be explored and the reduction in clutter that can be obtained by second harmonic imaging,both without and with contrast agents, is an excellent example of very recent progress. Newtransducer materials will probably result in greater sensitivity and better noise performanceand may make it possible to reduce exposures. Some of these materials may also be used inaffordable two-dimensional transducer arrays for three-dimensional image acquisition.

Although there are many similarities between Doppler (i.e., frequency or phase) and time-domain processing for obtaining information about blood flow and tissue motion, the twotechniques are often considered to be different and sometimes, in competition. Time-domainprocessing does have some advantages (e.g., there is no direct equivalent of ambiguity dueto the exceeding of the Nyquist limit), but it is computationally more demanding than theDoppler technique. As computing becomes more accessible and less expensive, however, this

Page 49: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 719

is becoming less of a consideration. In situations in which the greatest sensitivity is required,colour power imaging has an important low-noise advantage over colour velocity imaging,whether performed by Doppler or time-domain processing.

Three-dimensional imaging improves the perception of anatomical relationships, whetherby specialists reviewing complicated situations or by untrained observers with routine cases.There is scope for increasing the rate of three-dimensional image information acquisition bythe use of multiple-beam systems with parallel processing. Image segmentation remains aproblem, however, and this limits the opportunities for useful image display.

Amongst the specialized imaging methods, endoluminal techniques are already in routineuse. Synthetic aperture imaging may become useful, at least until two-dimensional arrayshave been further developed, and in microscanning applications. Computed tomography maybe used to provide information about tissue refraction and attenuation, for the improvementof traditional pulse–echo images. Elasticity imaging is a very promising technique; it may bedeveloped into quantitative telepalpation.

Although the primary contemporary role of ultrasonic imaging is in diagnosis, the methodalso has important applications in monitoring the progression and regression of disease,in some areas of screening, and in interventional procedures, both for localization and forguidance. Ultrasonic imaging is likely to become one of the preferred visualization techniquesin minimally invasive surgery, because of its high speed and ease of use.

The safety record of ultrasonic imaging is impeccable. There is no reason to suppose thatcontemporary techniques employ levels of exposure that could cause biological damage, butit is prudent to be cautious and further research is justified.

Acknowledgments

I am grateful to Dr H S Andrews for providing figures 10 and 21. The following figures arefrom product literature: Advanced Technology Laboratories, figures 7, 9, 15–18; Medison,figure 20; Endosonics, figure 23; and General Electric, figure 26.

References

Aanonsen S I, Barkve T, Naze Tjøtta J and Tjø S 1984J. Acoust. Soc. Am.75749–68Aarnik R G, Pathak S D, de la Rosette J J M C H,Debruyne F M J, Kim Y and Wijkstra H 1998Ultrasonics36635–42AIUM/NEMA 1992 Standard for the Display of Thermal and Acoustic Output Indices on Diagnostic Ultrasound

Equipment(Rockville: American Institute of Ultrasound in Medicine)Angelsen B A J1980IEEE Trans. Biomed. Engng2761–7Atkinson P and Berry M VJ. Phys. A: Math. Gen.7 1293–302Baker A C 1997Ultrasound Med. Biol.231083–8Ballantine H T, Hueter T F and Bolt R H 1954J. Acoust. Soc. Am.26581Bamber J C and Daft C 1986Ultrasonics2441–3Barber F E, Baker D W, Nation A W C, Strandness D E and Reid J M 1974IEEE Trans. Biomed. Engng21109–13Barnett S B, ter Haar G R, Ziskin M C, Nyborg W L, Maeda K and Bang J 1994Ultrasound Med. Biol.20205–18Barry C D, Allott C P, John N W, Mellor P M, Arundel P A, Thomson D A and Waterton J C 1997Ultrasound Med.

Biol. 231209–24Baum G and Greenwood I 1961NY State J. Med.614149–57Bohs L N, Friemel B H, Kisslo J, Harfe D T, Nightingale K R and Trahey G E 1995J. Am. Soc. Echocardiogr.8

915–23Bohs L N and Trahey G E 1991IEEE Trans. Biomed. Engng38280–6Bom N, Lancee C T, Honkeep J and Hugenholz P G 1971Biomed. Engng6 500–3, 508Bom N, Lancee C T and van Egmond F C 1972Ultrasonics1072–6Bom N, Lancee C T, van Zwieten G, Kloster F E and Roelandt J 1973Circulation481066–74Bonnefous O and Pesque P 1986Ultrason. Imag.8 75–85

Page 50: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

720 P N T Wells

Bouakaz A, de Jong N and Cachard C 1998Ultrasound Med. Biol.24469–72Bowen L, Gentilman R, Fiore D, Pham H, Serwatka W, Near C and Pazol B 1996Ferroelectrics187109–20Brandestini M A and Forster F K 1978IEEE Ultrason. Symp. Proc.(New York: IEEE) pp 348–52Brody W R and Meindl J D 1974IEEE Trans. Biomed. Engng21183–92Burkhardt C B, Grandchamp P-A and Hoffmann H 1974IEEE Trans. Sonics Ultrason.211–6Buschman W 1965Ultrasonics3 18–21Carson P L, Adler D D, Fowlkes J B, Harrist K and Rubin J 1992J. Ultrasound Med.11377–85Carstensen E L and Schwan H P 1959J. Acoust. Soc. Am.31305–11Chalana V, Linker D T, Haynor D and Kim Y 1996IEEE Trans. Med. Imag.15290–8Chen Y, Broschat S l and Flynn P J 1996Ultrason. Imag.18122–39Cosgrove D Oet al 1993Radiology18999–104Curry G R and White D N 1978Ultrasound Med. Biol.4 27–35de Jong N and Hoft L 1993Ultrasonics31175–81de Korte C L, van der Steen A F W, Cespedes E I and Pasterkamp G 1998Ultrasound Med. Biol.24401–8Deng J, Gardener J E, Rodeck C H and Lees W R 1996Ultrasound Med. Biol.22979–86Desch C H, Sproule D O and Dawson W J 1946J. Iron Steel Inst.153319–27Detmer P R, Bashein G, Hodges T, Beach K W, Filer E P, Burns D H and Strandness D E 1994Ultrasound Med. Biol.

20923–36Donald I, MacVicar J and Brown T G 1958Lancet1 118–94Doppler C 1843Abh. K.-bohm. Ges.2 465–82Duck F A 1990Physical Properties of Tissues(London: Academic)Duck F A and Martin K 1991Phys. Med. Biol.161423–32Dussik K T, Dussik F and Wyt L 1947Wien. Med. Wschr.97425–9Eggleton R C, Feigenbaum H, Johnson K W, Weyman A E, Dillon J C and Chang S 1975Ultrasound in Medicineed

D White (New York: Plenum) pp 385–95Embree P M and O’Brien W D 1985IEEE Ultrason. Symp. Proc.(New York: IEEE) pp 963–6Erbel R, Roth T, Koch L, Ge J, Gorge G, Serruys P W, Bom N, Lancee C T and Roelandt J 1997Minimally Invasive

Therapy and Allied Technol.6 195–8Eyers M K, Brandestini M, Phillips D J and Baker D W 1981Ultrasound Med. Biol.7 21–31Fairbank W M and Scully M O 1977IEEE Trans. Biomed. Engng24107–10Farmery M J and Whittingham T A 1978Ultrasound Med. Biol.3 373–9Finette S 1987IEEE Trans. Ultrason. Ferroelectr. Freq. Control34283–92Firestone F A 1946J. Acoust. Soc. Am.17287–99Fish P J 1981Recent Advances in Ultrasound Diagnosisvol 3, ed A Kurjak and A Kratochwil (Amsterdam: Excerpta

Medica) pp 110–15Fishman E K, Magid D, Ney D R, Chaney E L, Pizer S M, Rosenman J G, Levin D N, Vannier M W, Kuhlman J E

and Robertson D D 1991Radiology181321–37Follett D H 1986Physics in Medical Ultrasounded J A Evans(London: IPSM) pp 78–84Forsberg F, Goldberg B B, Liu J-B, Merton D A and Rawool N M 1996J. Ultrasound Med.15853–60Frinking P J A and deJong N 1998Ultrasound Med. Biol.24523–33Fry W J and Fry R B 1954J. Acoust. Soc. Am.2624–317Gao L, Parker K J, Lerner R M and Levinson S F 1996Ultrasound Med. Biol.22959–77Gardiner W M and Fox M D 1989Radiology172866–8Gehlbach S M and Sommer F G 1987Ultrason. Imag.9 92–105Gramiak R, Shah P M and Kramer D H 1969Radiology92939–48Greenleaf J F, Johnson S A, Lee S L, Herman G T and Wood E H 1974Acoustical Holographyvol 5, ed P S Green

(New York: Plenum) pp 591–603Hadjicostis A N, Hottinger C F, Rosen J J and Wells P N T1984Ferroelectrics60107–26Hahn E L 1960J. Geophys. Res.65776–7Halliwell M, Key H, Jenkins D, Jackson P C and Wells P N T1989Br. J. Radiol.62824–9Harris R A, Follett D H, Halliwell M and Wells N T 1991Ultrasound Med. Biol.17547–58Hasegawa T and Yoshioka K 1969J. Acoust. Soc. Am.461139-43Holm H H, Kritorven J K, Pederson J F, Hancke S and Northered A 1975Ultrasound Med. Biol.2 19–24Holmes J H, Wright W, Meyer E P, Posakony G T and Howry D H 1965Am. J. Med. Electron.4 147–52Hoskins P R 1997Ultrasound Med. Biol.23889–97Hoskins P R and McDicken W N 1997Br. J. Radiol.70878–90Hounsfield G N 1973Br. J. Radiol.461016–22Howry D H and Bliss W R 1952J. Lab. Clin. Med.40579–92

Page 51: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

Ultrasonic imaging of the human body 721

Hueter T F and Bolt R H 1951J. Acoust. Soc. Am.23160–7Hughes D I and Duck F A 1997Ultrasound Med. Biol.23651–64Insana M F, Wagner R F, Brown D G and Hall T J 1990J. Acoust. Soc. Am.87179–91Jaffe B, Roth R S and Marzullo S 1955J. Res. Nat. Bur. Stand.55239–54Jago J R and Whittingham T A 1991Phys. Med. Biol.361515–27Kallel F, Ophir J, Magee K and Krouskop T 1998Ultrasound Med. Biol.24409–25Kasai C, Namekawa K, Koyano A and Omoto R 1985IEEE Trans. Sonics Ultrason.32485–64Kessler L W 1974J. Acoust. Soc. Am.55909–18Koch C H 1996Ultrasonics34687–9Kossoff G 1972Ultrasonics10221–7Kremkau F W, Gramiak R, Carstensen E L, Shah P M and Kramer D H 1970Am. J. Roentg.110177–83Lemons R A and Quate C F 1974Appl. Phys. Lett.25251–3Levinson S F, Shinagawa M and Sato T 1995J. Biomech.281145–54Lewin P A 1981Ultrasonics19213–16Li W, Bom N and van Egmond F C 1995J. Vasc. Invest.1 57–61Lypacewicz G and Hill C R 1974Ultrasound Med. Biol.1 287–9Machi J, Sigel B, Beitler J C, Coelo J C U andJustin J R 1983J. Clin. Ultrasound113–10McDicken W N, Bruft K and Patton J 1974Ultrasonics12269–72McDicken W N, Lindsay M and Robertson D A R 1972Br. J. Radiol.4570–1McDicken W N, Sutherland G R, Moran C M and Gordon L N 1992Ultrasound Med. Biol.18651–4McGahan J P and Goldberg B B (ed) 1997Diagnostic Ultrasound(Philadelphia, PA: Lippincott-Raven)Minnaert M 1933Phil. Mag.16235–48Mo L Y L and Cobbold R S C1986IEEE Trans. Biomed. Engng3320–7Mosfeghi M and Waag R C 1988Ultrasound Med. Biol.14415–28Nelson T R and Pretorius D H 1998Ultrasound Med. Biol.241243–70Nguyen N T, Lethiecq M, Karlsson B and Patat F 1996Ultrasonics34669–75Nippa J H, Hokanson D E, Lee D R, Sommer D S and Strandness De 1975IEEE Trans. Sonics Ultrason.22340–6O’Donnell M 1983IEEE Trans. Sonics Ultrason.3026–36O’Donnell M, Eberle M J, Stephens D N, Litzza J A, San Vicente K and Shapo B M 1997 IEEE Trans. Ultrason.

Ferroelectr. Freq. Control44714–21O’Donnell M, Skovoroda A R, Shapo B M and Emelianov S Y 1994IEEE Trans. Ultrason. Ferroelect. Freq. Contr.

41314–25Ophir J, Cespedes I, Ponnekanti H, Yazdi Y and Li X 1991Ultrason. Imag.13111–34Patzold J, Krause W, Kresse H and Soldner R 1970IEEE Trans. Biomed. Engng17263–5Porter T R, Shouping L, Kriesfield D and Ambruster R W 1997J. Am. Coll. Cardiol.29791–9Preston R C, Bacon D R, Livett A J and Rajendran K J 1983J. Phys. E: Sci. Instrum.16786–96Ramachandran G N and Lakshminarayanan A V 1971Proc. Natl Acad. Sci., USA682236–40Redman J D, Walton W P, Fleming J E and Hall A J 1969Ultrasonics7 26–9Reid J M and Baker D W 1971Ultrasonographia Medicavol I, ed J Bock and K Ossonig (Vienna: Wiener Medizischen

Akademie) pp 109–20Reid J M and Spencer M P 1972Science1761235–6Ries L L and Smith S W 1995Ultrason. Imag.17227–47Ritchie C J, Edwards W S, Mack L A, Cyr Dr and Kim Y 1996Ultrasound Med. Biol.22277–86Rooney J A 1973Ultrasound Med. Biol.1 13–16Rubin J M, Bude R O, Carson P L, Bree R L and Adler R S 1994Radiology190853–6Schrope B A, Newhouse V L and Uhlendorf V 1992Ultrason. Imag.14134–58Schuette W H, Shawker T H and Whitehouse W C 1978J. Clin. Ultrasound6 16–18Sehgal C M and Arger P H 1997J. Ultrasound Med.16471–9Seyed-Bolorforosh M S 1996Ultrasonics34135–8Shankar P M, Dala Krisna P and Newhouse V L 1998Ultrasound Med. Biol.24395–9Shattuck D P, Weinshenker M D, Smith S W and von Ramm O T 1984J. Acoust. Soc. Am.751273–82Smith S W, Pavy H G and von Ramm O T 1991IEEE Trans. Ultrason. Ferroelectr. Freq. Control38100–8Smith S W, Trahey G E and van Ramm O T 1986Ultrasound Med. Biol.12229–43Somer J C 1968Ultrasonics6 153–9Staudenraus J and Eisenmenger W 1993Ultrasonics31267–73ten Hoff H, Gussenhoven E J, Korbijn A, Mastik F, Lancee C T and Bom N 1995Eur. J. Ultrasound2 227–37Thomas J D and Rubin D N 1998J. Am. Soc. Echocardiogr.11803–8Torr G R and Watmough D J 1977Phys. Med. Biol.22444–50

Page 52: Ultrasonic imaging of the human body - Semantic …...Ultrasonic imaging of the human body 673 1. Introduction More than one out of every four medical diagnostic imaging studies in

722 P N T Wells

Trahey G E, Allison J W and von Ramm Ot 1987IEEE Trans. Biomed. Engng34965–7Vilkomerson D H R, Mezrich R S and Etzold K F 1977Acoustical Holographyvol 7, ed L W Kessler (New York:

Plenum) pp 87–101von Ramm O T and Smith S W 1983IEEE Trans. Biomed. Engng30438–52von Ramm O T, Smith S W and Pavy H G 1991IEEE Trans. Ultrason. Ferroelectr. Freq. Control38109–15Wagner R F, Smith S W, Sandrik J M and Lopez H 1983IEEE Trans. Sonics Ultrason.30156–63Walker C L, Foster F S and Plewes D B 1998Ultrasound Med. Biol.24137–42Walker W F, Friemel B H, Laurence N B and Trahey G E 1993Proc. IEEE Ultrason. Symp.873–7Wells P N T1966Wld. Med. Electron.4 272–7——1969Med. Biol. Engng7 641–52——1970Rep. Prog. Phys.3345–99——1975Ultrasound Med. Biol.1 369–76——1977Biomedical Ultrasonics(London: Academic)——1986Br. J. Radiol.591143–51——1995J. Vasc. Invest.1 38–43Wells P N T,Bullen M A, Follett D H, Freundlich H E and James J A 1963Ultrasonics1 106–10Wells P N T andHalliwell M 1981Ultrasonics19225–9Wells P N T,Halliwell M, Skidmore R, Webb A J and Woodcock J P 1977Ultrasonics15231–6WFUMB 1992Ultrasound Med. Biol.18731——1997World Federation for Ultrasound in Medicine and Biology News4 no 2 (1997Ultrasound Med. Biol.23

following p 974)Whittingham T A 1976Ultrasonics1429–33Wild J M and Reid J M 1952Am. J. Path.28839–61Ylitalo J 1996Eur. J. Ultrasound3 277–81Ylitalo J and Ermert H 1994IEEE Trans. Ultrason. Ferroelectr. Freq. Control41333–9Yuan Y-W and Shung K K 1989J. Ultrasound Med.8 425–34Zheng W and Newhouse V L 1998Ultrasound Med. Biol.24513–22Zimmer Y, Tepper R and Akselrod S 1996Ultrasound Med. Biol.221183–9Ziskin M C, Bonakdarpour A, Weinstein D P and Lynch P R 1972Invest. Radiol.7 500–5