Contents 1. 1.1. 2. Preliminaries 1 Method of Teaching 1 What is ultrasound imaging? 2 2.1. What do we mean by ultrasound? 2 2.2. Uses of ultrasound imaging 3 3. A-Mode (F51–68, C123–135, W343–347) 3 3.1. Uses 3 3.2. What causes the reflections? (F 21, C 26–36, W 325) 4 3.3. What is the significance of all the reflection coefficients? 5 3.4. What other aspects of wave propagation are important? 7 3.4.1. Scattering (F 22–25, C 189, W 324) 3.4.2. Absorption (F 23–25, C 214, W 323–324) 3.4.3. Diffraction (F 28–39, C 89–104, SW 328–333) 3.4.4. Worked Examples 3.5. 3.6. 4. 7 8 9 10 How is it all achieved in practice? 11 3.5.1. Master Clock (PRF Generator) 3.5.2. Transmitter/Transducer/Receiver 3.5.3. Time Gain Compensation (TGC) 3.5.4. Demodulator 12 13 13 14 What is the right ultrasound frequency to use? 15 3.6.1. Transducer Resonance 3.6.2. Transducer response characteristic in the frequency domain 3.6.3. Transducer response characteristic and the time domain signal 3.6.4. Q-values and pulse lengths 3.6.5. Resolution and Frequency (F184–185, C82–89, 104–112) 3.6.6. Attenuation and frequency 15 17 17 18 19 20 B-Mode (F69–96, C135–158, W348–351) 20 4.1. Introduction 20 4.2. What do we need to add to an A-scanner to turn it into a B-scanner? 21 4.3. The Co-ordinate Generator 21 4.4. Compression and Amplifier 22 The Beam-Steering Device 23 4.5. 4.5.1. Static B-Scanners 4.5.2. Real-Time Mechanical Scanners (F83–84, C144–146) 4.5.3. Electronic Steering — Transducer Arrays (F89–94, C146–158, W340–343) i 23 24 25 5. The Theory of Diffraction 25 5.1. Making Huygens’ Wavelets Quantitative 25 5.2. The plane piston transducer 28 Two very small transducers Young’s Slits 29 5.4. 30 The linear array transducer Electronic beam steering re-visited 31 5.6. Lateral Resolution, Diffraction and Focussing 33 Artifacts … or … Why do ultrasound images look so awful? (F179–189) 37 5.7. 6. 5.7.1. Misregistration 5.7.3. Speckle 5.7.5. If ultrasound images are so bad, why use them? 37 39 40 Doppler Ultrasound (F117–164, C186–212, W351–358) 40 6.1. Introduction 40 6.2. The Doppler Equations 40 How do we put all this into practice? 42 Second transducer for reception 6.3.2. Doppler demodulator 6.3.3. Frequency meter / spectrum analyser 6.3.4. Real-time display Problems with CW Doppler 43 44 44 45 45 6.3. 6.4. Duplex Scanners 45 6.5. Limitations of Doppler Ultrasound 46 7. Bio-effects of Ultrasound (F201–212, C213–218, W375–377) 46 7.1. Introduction 46 7.2. Heating 46 7.3. Radiation Pressure 48 7.4. Acoustic Streaming 49 7.5. Cavitation 49 7.6. Therapeutic Ultrasound 51 7.6.1. Physiotherapy 7.6.2. Surgical Uses 7.7. 51 51 And finally ... is it safe? 51 ii 1. Preliminaries 1.1. Method of Teaching This is the fourth year of this course. Things in this handout have been improved continually, as a result of the response of the previous year’s students, so I will be very happy to hear any feedback you might have. Not everything you need to know is in the handouts. There are gaps to fill in with extra notes which I give in lectures. For the exam you need to know these as well. There is an extra Appendix handout, which contains derivations and other background material. None of the derivations in the Appendix will be examined, but you will be expected to understand the definitions of various acoustical quantitites such as acoustic pressure and characteristic acoustic impedance. The Appendix will be useful for you to look at it so you can understand where the results come from. PwA students are also taking the course Ultrasonics and will have some background in wave propagation which the PMP students do not. The exam is designed so that this should not matter. Recommended books: Diagnostic Medical Ultrasound, Peter Fish, John Wiley and Sons 1990, ISBN 0-47192651-5 (about £20) — abbreviation in notes F followed by a page number. The Physics of Medical Imaging, Ed. S. Webb, Adam Hilger 1990, 361-1 (about £20) — abbreviation W. ISBN 0-85274- Ultrasonic Bioinstrumentation, Douglas Christensen, John Wiley and Sons, 1988, ISBN 0-471-60496-8 (about £80, try library) — abbreviation C. Introduction 1 2. What is ultrasound imaging? Ultrasound imaging operates on the same principle as radar and sonar and is similar to the echo-location method bats use to navigate. An emitter sends out pulses of sound. These bounce of objects and the returned echoes give us information about the object. Reflector Emitted pulse Transducer 0 50 100 150 c 200 0 c 5 0 1 0 0 1 5 0 2 0 0 Lower amplitude reflected pulse d Figure 1: Basic principle of ultrasound imaging The fundamental equation of ultrasound is d ct 2 [1] d = distance of the reflecting object from the source/detector of ultrasound; c = speed of the ultrasound; t = round-trip time of the pulse, from emission to reception. 2.1. What do we mean by ultrasound? Acoustic waves with frequencies above those which can be detected by the human ear. In practice, 20 kHz < f < 200 MHz. An acoustic wave is a propagating disturbance in a medium, caused by local pressure changes at a transducer. Wavefronts Rarefaction Compression c The molecules of the medium oscillate about their resting (equilibrium) positions, giving rise to a longitudinal waves. Figure 2: Longitudinal waves in gas 2 c 1540 m/s 6.5 ms/cm in most body tissues. = c / f = 1.5 mm at 1 MHzGases and liquids support only longitudinal waves; solids support transverse waves as well, but these are rapidly attenuated for non-rigid, “soft” solids. 2.2. Uses of ultrasound imaging Most widespread use is medical imagingNon-invasive, low riskObstetrics, abdominal problems, measurement of blood flow and detection of constrictions in arteries and veins.Also used in non-destructive testing in industry: e.g., cracks in structures. S o n a r , Figure 3: A typical obstetric ultrasound scan underwater imaging (e.g., in submarine echo-location devices). 3. A-Mode (F51–68, C123–135, W343–347) Simplest form of ultrasound instrumentPulses of ultrasound in a thin beam are emitted from a transducer into the body and encounter interfaces between different organs.Some of the sound energy is reflected at each stage and some continues through to be reflected in turn by deeper organs.The returning pulses are detected by the transducer and the amplitude of the signal is displayed on an oscilloscope. If the time-base of the scope is constant, then the distance across the screen corresponds to the depth of the object producing the echo, according to Eq. [1]. 3.1. Uses Gives information very quicklyMinimum of sophisticated apparatusWeakness is that this information is one-dimensional — i.e., along the line of the beam propagation.Nowadays, this mode has been largely superseded by B-mode (see later). 3 A-mode still finds uses in ophthalmology, where the simple structure of the eye makes it relatively easy to interpret the echoes and where what is required are straightforward but accurate measurements of, for example, distance from the lens to the retina. Even this very primitive instrument is not as straightforward as it might seem. To understand why, we need to look at a number of principles of physics, engineering and signal processing. 3.2. What causes the reflections? (F 21, C 26–36, W 325) Reflections occur when the incident wave encounters a boundary between two materials with different acoustic impedances.Acoustic impedance Z is the material property which relates pressure changes p (in excess of atmospheric) to the vibrational velocity u of the particles in the medium. p Zu [2] If we are looking at a single plane wave through a substance with density and speed of sound c, then Z = c.See the Appendix for a more lengthy description. When an incident plane wave, travelling through a medium with acoustic impedance Z1 hits a boundary with a second material of impedance Z2 at an angle i to the normal, there is a reflected wave at an angle r, and a transmitted wave, refracted at an angle t . The following important laws apply: Law 1 The direction vectors of the incident wave, the reflected wave and the transmitted wave are all in the same plane. Law 2 The angle of incidence is equal to the angle of reflection. i = r Law 3 The angle of incidence and the angle of transmission (refraction) are related by Snell's Law: sin i c1 Z / 1 1 sin t c2 Z2 / 2 , [3] where the Z1 and Z2 are the characteristic acoustic impedances of materials 1 and 2 (see Appendix), c1 and c2 are the speeds of sound and 1 and 2 are the (equilibrium) densities. 4 Law 4 The pressure amplitudes pr and pt of the reflected and transmitted waves are related to that of the incident wave pi by pr Z2 cos i Z1cos t pi ; Z2 cos i Z1cos t pt 2 Z2 cos i pi . Z2 cos i Z1cos t [4] Note that pr and pt are linked by pi+pr=pt. At normal incidence, this becomes pr Z2 Z1 pi ; Z2 Z1 pt 2 Z2 pi Z2 Z1 [5] For those interested, the derivation of these laws is given in the Appendix. You do not need to know the derivation for the exam, but you do need to know Laws 1–4. 3.3. What is the significance of all the reflection coefficients? (i) Refraction effects are important. If the beam can be turned through an angle, then we can not guarantee the path of the return signal. We don't know exactly where the echo has come from. It might theoretically be possible to calculate all these effects, but the phenomenon certainly complicates things. In practice, simple scanners ignore these effects and errors are possible. (ii) Too little reflection is bad. pr / pi 0 Useful images occur only where there is a difference in acoustic impedance. Tissues with strikingly different properties in other respects may have similar acoustic impedances. From Fig. 5, you can see that there is virtually no reflection at a transition from liver to spleen and so the two tissues will not be delineated one from the other. Transducer Z1 Normal i Z2 Image t Interface Figure 4: Displacement of image due to refraction. (Here Z2 < Z1.) 5 Reflector Tendon/Fat Water/Muscle 0 -10 -20 -30 Liver/Spleen -40 -50 dB (pr/ pi)2 1 0.1 0.01 Bone/Soft Tissue Air/Solid Air/Liquid Figure 5: 10-3 10-4 10-5 Muscle/Liver Lens/Vitreous or Aqueous Humour Reflection coefficients at various tissue boundaries. Note that these are power reflection coefficients (see later). (iii) Too much reflection is bad. pr / pi 1 If the difference in acoustic impedance is too high, then virtually all the incident ultrasound will be reflected. This means that the boundary is opaque to ultrasound. The organ in question will show up very brightly, but we won't be able to see through it to find out what is underneath.No ultrasound images of brain in vivo; skull reflects ultrasound Images of the heart have to be taken “round” the ribs, which are also opaque. Finding the right “window” into the body is important. (iv) The ultrasound transducer must be “coupled” to the body using a special gel. Before an ultrasound scan, a thin layer of gel is smeared onto the skin. Why? Answer: (F p.45–47) The material from which transducers are made has a very different acoustic impedance Ztransducer to that of the body Ztissue and more importantly that of air Zair. These large “mis-matches” between Ztransducer and Ztissue and between Ztransducer and Zair mean that the refelection coefficients at these interfaces are close to 1. Little of the signal gets through gets through at a transducer-tissue boundary (pr/pi0.86) and virtually none at a transducer-air boundary (pt/pi 0.9997). By applying the coupling gel, we exclude all air from the region between probe and body. This means that the worst case scenario of reflection from a 6 transducer-air boundary is avoided. The reflection coefficient is still high (0.86), but imaging is possible. Some manufacturers use a trick called impedance matching to increase the amount of transmitted radiation through a transducer-tissue interface. Sec. A9 explains mathematically how this works. Inside the probe, there is a matching layer of thickness /4 between the transducer and the tissue. The acoustic impedance of the matching material is approximately: Zmatch Ztransducer Ztissue [6] The technique has analogues in optics (“blooming” of lenses), electronics (coaxial transmission lines) and quantum mechanics (scattering of particles by potential wells). Note that this technique is not suitable in all cases and, in particular, a /4 layer will match completely only a single frequency of ultrasound. 3.4. What other aspects of wave propagation are important? 3.4.1. Scattering (F 22–25, C 189, W 324) Formulae in Sec. 3.2 strictly valid only for an infinite plane reflecting surfaceIn the body, there are many structures which are much smaller than this (e.g., lung tissue (a) (b) Transducer Soft Tissue “Matching” gel Transducer pi Soft Tissue pi pt pt pr pr |pt| << |pr| |pt| >> |pr| Figure 6: (a) A large degree of reflection occurs at the interface between the ultrasound transducer and soft tissue. (b) If the correct thickness of an appropriate material is built into the probe, much improved transmission can be obtained. Note that there is still a thin gel layer (not shown) between the “matching” layer inside the probe and the tissue. This has approximately the same acoustic impedance as the soft tissue and is used to exclude air. 7 is a fine network of air-filled tubes).These give rise to a whole series of interfaces, at random orientations, and the reflections from these scatter the incident wave.A typical 1 MHz ultrasound beam does not penetrate lung tissue. At a smaller scale where d<<(e.g., red blood cells), Rayleigh Scattering occurs: The degree of scattering varies as f4 This means that low frequency ultrasound penetrates tissue better. The distribution of scattered radiation is approx. uniform (isotropic re-radiation). The exact shape of the particle does not matter very much. 3.4.2. Absorption (F 23–25, C 214, W 323–324) A phenomenon by which organised vibrations of molecules (i.e., ultrasound) are transformed into disorganised, random motion. Acoustic energy Heat The mechanisms for this transfer include fluid viscosity, molecular excitations and chemical changes.This is still largely uninvestigated for most tissues. Since biological tissues are so complex, this is not surprising. It is difficult to measure the proportion of energy loss which occurs by scattering and the proportion lost by absorption. Assuming that the tissue is uniform, we find that the reduction in amplitude of the ultrasound over a small distance x, is a constant fraction of the amplitude at the point x, wherever that is along the path. p0 (p0 )x . [7] In the limit x 0, this gives the differential equation dp0 dx p0 , [8] . [9] which integrates to p0 p0 ( 0 ) e x We can do the same thing for the scattering effect, so that overall 8 p0 ( x) p0 (0) exp[( scatter absorb )x] . [10] This also applies to the peak oscillation velocity u0 and the amplitude of displacement a0 of the particles. Notes: (i) Attenuation is approximately proportional to frequency, so that the depth of penetration goes down as f rises. (ii) Instead of using amplitude, attenuation is often measured in terms of a reduction in the power density transported by the wave. Consider the units of pu, where u is the particle vibration velocity: Pressure Velocity = N/m2 m/s = (Nm)s-1/ m2 = W/m2 = Power/unit area I.e., pu represents the power being transported by the ultrasound through a unit area of the tissue normal to the direction of propagation. It is often also called the intensity of the ultrasound and is represented by the symbol I. If we look at the power (intensity) attenuation, we see that I pu Re[ p0 e i (t kx ) ] . Re[u0 e i (t kx ) ] p0 u0 cos 2 (t kx ) I0 cos 2 (t kx ) . [11] Now p0(x) = p0(0) e-x and similarly for u0. Hence I 0 ( x ) I 0 (0) e 2x . [12] The power density transported decays twice as quickly as the vibration amplitude. (iii) Attenuation is often measured on a decibel scale, where I( x) Attenuation in dB 10 log10 I (0) . [13] 3.4.3. Diffraction (F 28–39, C 89–104, SW 328–333) Huygens’ Principle states that each point on a wavefront can be regarded as acting like a secondary source, emitting spherical wavelets. The new overall wave is found by summing the contributions from all the individual wavelets. Interference between these wavelets gives rise to diffraction effects. 9 Incident wavefronts Line of constructive interference Secondary sources Wavelet Figure 7: Diffraction effects Diffraction becomes significant when the apparatus dimensions and objects examined become comparable with the radiation wavelength. Thus acoustic diffraction (0.1mm) is a much more significant effect than optical diffraction (500nm) for biological tissues. 3.4.4. Worked Examples A 1 MHz US wave with initial intensity 100 mW/cm2 (RMS) is travelling through fat. (We shall assume initially that we have a semi-infinite slab so that there is no reflected component.) Calculate: (a) the initial peak pressure, (b) the initial maximum velocity of oscillation of the particles, (c) the initial maximum displacement. Now suppose that, the beam hits a barrier with muscle 3cm from the starting point. Calculate: (d) the intensity of the reflected beam. The tissue properties are: fat = 940 kg/m3, c = 1480 m/s, (amplitude) = 0.07 cm-1; muscle = 1070 kg/m3, c = 1566 m/s, (amplitude) = 0.15 cm-1. (a) Peak Pressure We have I = p u = p2/Z = p02 cos2t / Z We are given the RMS intensity, i.e., IRMS = p02 <cos2t> / Z = p02 / 2Z For a plane wave, Z =c p0 = (2cIRMS) = (2 940 kgm-3 1480 ms-1 1000 Wm-2)1/2 = (2.78 109 kg2m-2s-4)1/2 = 5.27 104 kgms-2m-2 = 5.27 104 N/m2 or about 1/2 atmosphere. Remember, this is the excess pressure above atmospheric. This calculation gives the peak pressure. The RMS pressure is p0/2. (b) Peak Velocity p0 Zu0 u0 5.27 10 4 kg m -1s-2 0.038 m / s 3.8 cm / s . 1.39 10 6 kg m -3 . m s-1 10 (c) Peak Displacement q , u u0 e i (t kx ) t where q is the displacement of the particle from its mean position. Integrating this expression gives u q 0 e i (t kx ) , i where the i in the pre-multiplying factor indicates that q is 90 out of phase with u (1/i can be re-expressed as e-i2 and so the whole phase angle is ei(t - kx - ). q0 = u0 / and for 1 MHz ultrasound, = 2f = 2 106 rad s–1. q0 = 3.8 10 2 6.0 10 9 6.0 nm . 2 106 (d) Intensity of Reflected Beam By the time it reaches the interface, the incident beam has peak intensity I (3 cm) I (0) e 2x 100 mW/cm 2 e 2 0.07cm -1 3 cm 65.7 mW/cm 2 . This is equivalent to an attenuation of 10 log10 0.657 182 . dB . The reflection coefficient for amplitude of pressure fluctuations is R pr Z2 Z1 1.68 1.39 0.0945 10.2 dB , pi Z2 Z1 1.68 1.39 where Z1 = c for fat and Z2 = c for muscle. Since both pressure and velocity are reduced by this factor, the reflected beam has its intensity reduced by (0.0945)2 20.4 dB. the intensity of the reflected wave at the interface is 65.7 mW/cm2 (0.0945)2 = 0.59 mW/cm2 to 2 s.f. So, 1/3 of the incident power is dissipated by attenuation processes in the fat before it reaches the interface. Of the remaining 2/3, most of the power is transmitted, because the acoustic impedances of the two tissues are quite similar. 3.5. How is it all achieved in practice? Practical design of an A-mode scanner (F51–68, C124–135, W343–348) Fig. 8 shows the block diagram of a practical scanner. The new additions as compared with the simple diagram are concerned with the practical problems one discovers when one tries to use the reflected ultrasound signal. 11 V t PRF generator Reflecting objects Pulse generator Transducer 0 50 100 150 200 0 5 0 1 0 0 1 5 0 2 0 0 Protection circuit 70-80 dB Variable gain amplifier (TGC) TGC generator 40-50 dB V(dB) t Demodulator y Display ’scope timebase x V t Figure 8: Block diagram of a practical A-scanner. N.B. Not all A-mode scanners include a demodulator. The dynamic range values at each stage are approximate and refer to the power range in the signal. Take the square root (i.e., halve the dB value) for the corresponding amplitude ranges. How can the same probe both transmit pulses and receive the echoes? How do we cope with the signal attenuation by tissue?How do we display the signal?Master Clock (PRF Generator) This synchronises the various parts of the scanner (e.g., transmitter, receiver, oscilloscope timebase) so that each is triggered to act at the correct time. PRF stands for pulse repetition frequency, the frequency at which clock pulses occur and at which ultrasound pulses are sent out into the sample. 12 3.5.2. Transmitter/Transducer/Receiver On the leading edge of each clock pulse, either a momentary voltage step, or a short sinusoidal burst of voltage is applied to the transducer. The transmitter which performs this must have a short rise time, i.e., it must be able to go from zero to its maximum voltage (100–200 V) very quickly (typically <25ns), in order to produce ultrasound pulses with very high frequency components. 3.5.3. Time Gain Compensation (TGC) Problem: Ultrasound is attenuated as it passes through tissue. So, even for the same type of reflector, the signal is less for deeper objects. This effect is very significant. The worked example (Sec. 3.4.4) showed a typical value of 0.15 cm-1, so on a typical return trip of 10cm, the signal is reduced by e-0.1510 = 0.22 compared with reflections coming straight from the skin surface. Solution: Amplify the later-arriving signals (i.e., the ones from deeper in the tissue) more. I.e., change the receiver gain with time to compensate for the echo attenuation.Achieved by making the gain of the amplifier dependent on a control voltage. This input voltage is changed by the TGC unit.Because of the logarithmic nature of the decrease in signal, the TGC should increase the gain a certain number of dB each ms. Worked Example An ultrasound beam propagates in uniform liver tissue with = 0.15 cm-1. If the speed of sound in the tissue is c = 1540 m/s, what should be the rate of gain increase by the TGC? = 0.15 cm-1 is the attenuation coefficient for amplitude; the power attenuation is double this. Amplifiers are specified in terms of power, so 2 = 0.3 cm-1 is what we want. In terms of dB, we have 10 log10 ( e-0.3 ) = 1.3 dB/cm to 2 s.f. So we want a gain increase of 1.3 dB for each cm of travel to cancel this out. Now 1540 m 1 s 1 cm 1.3 dB 1 ms . This means that the require TGC rate is 154 1 ms 200 dB / ms. Clearly, a specialised amplifier is needed. 154 In practice, tissue type varies with depth and the situation is more complicated. The user is given a range of controls to vary the TGC. The rate of increase of gain (i.e., d2G/dt2 ) varies with time and hence depth. This is not an exact science! Notice, too, that by “tweaking” the time-gain controls to get a better images, we lose the information provided by the attenuation coefficient. By using this compensation we are “ignoring” the physics of the situation. The fact that one might not be able to see a particular boundary tells us something about the properties of that boundary. 13 3.5.4. Demodulator At the output of the compression amplifier, the echo signal mirrors that of the pulse, i.e., it oscillates at the ultrasound frequency of several MHz.The display is much easier to understand if this high frequency modulation is removed.Another way of describing demodulation is to say we want to change a signal oscillating at a high frequency Let us multiply an original signal S = cos hight by cos 0t. A useful trigonometry theory states that: cos high t cos 0 t 1 cos( high t 0 t ) cos( high t 0 t ) . 2 [14] Variety of simple pre-programmed shapes for increasing the gain Gain / dB 60 40 Vclock 20 0 Vin 0 2 4 6 8 10 Depth / cm Vcontrol 1/PRF t Vcontrol Set of levels all adjustable by operator for more flexibility Variable gain amplifier (TGC) t Gain / dB 60 40 Vout 20 0 0 2 4 6 8 Vout = G( Vcontrol ) . Vin 10 Depth / cm Figure 9: Principle of operation of time gain compensation (TGC) 14 Original Signal Reference Signal Low-pass filter Result at difference frequency Figure 10: Principle of demodulation by multiplication by a reference frequency Now if we use a low pass filter, , we can cut out the first term leaving a signal of the form cos( high t 0 t ) cos( t ) . [15] Any echo can be made up of a sum of such terms. The multiplying and filtering process simply shifts all the terms down in (angular) frequency by 0.. This gives us the envelope of the signal, as in Fig. 10. 3.6. What is the right ultrasound frequency to use? No universal “right” frequency Ultrasound frequency chosen is determined by a number of factors: scan time for cos t images, resolution, the length of the Vultrasound pulses and the transducer design.To ~ understand how these interact, we start with a simple explanation of how transducers work. 0 Transducer Desired ultrasound insonating sample Ultrasound coming from back of transducer 3.6.1. Transducer Resonance Transducers exploit the piezo-electric effect.When a voltage is applied across certain types of crystal, the crystalline structure of the crystal is distorted and a Interfering US in transducer ~ /2 for fundamental 15 Figure 11: Principle of operation of a transducer change in size occurs. If the crystal is made such that the change in size happens perpendicular to the faces. Apply voltage perpendicular to faces oscillating at the ultrasound frequency. These move in and out generating pressure waves.Ultrasound waves both outside and inside the crystal Waves inside the crystal are reflected off the opposite face and return to interfere with the original vibration.Waves may come back either in phase with the original vibration, out of phase, or somewhere in between. Only if the waves reinforce each Frequency Domain Time Domain Observed Vibration Spectral Response A0 Exponential envelope R0 Q = 0 / 2 R0/2 A0 / e 2 0 0 50 100 150 200 00 0 50 50 00 / Q 10 100 150 150 200 200 T=2/0 Fig. 12: Ideal transducer characteristics in the frequency and time domain other will we get an ultrasound signal. Relative phases of the original vibration and the reflected wave depend on the wavelength of the ultrasound in the crystal. So the amplitude of the signal generated by the transducer depends on the frequency of the oscillating voltage. Transducers exhibit resonance properties.Because of the relative Z values for the transducer and the external medium, we require the faces of the transducer to be nodes (almost). This might at first seem counter intuitive.For simple cases, the width of the transducer must be an odd multiple of Fundamental frequency and . In more realistic situations, we want all the radiation to exit from the front of the transducer. This means that a backing material is placed at the rear face to absorb backward travelling ultrasound. This changes the boundary conditions for the wave equation and hence the width of the transducer at the fundamental frequency. 16 t 3.6.2. Transducer response characteristic in the frequency domain Around each of its resonant frequencies, the transducer will have a response curve like that shown in Fig. 12. Typically, the curve will have a Lorentzian form such as R R0 1 ( 0 ) 2 / ( ) 2 , [16] is the (half) width of the response function at half height. We define a quantity called the quality factor or “Q-value” of the transducer by Q 0 Resonant frequency Linewidth 2 . [17] A high-Q circuit has a very sharply peaked resonance, while a low Q corresponds to a broad line. The size of Q depends on the losses from the resonant circuit containing the transducer. Since transducers are generally constructed out of materials having low internal loss, it follows that the largest contributor to losses is radiation of power into the surrounding material, i.e., tissue in a medical instrument. In cases where we want very low losses, such as in highly stable quartz oscillators, the transducer is bounded by air or a vacuum and the Q can be up to 106. For pulsed ultrasound, where the aim is to radiate energy into the tissue, Q 5–15. 3.6.3. Transducer response characteristic and the time domain signal How does the above response characteristic relate to the actual time variation of the vibration of the crystal when we create a pulse of ultrasound? It can be shown (Fourier Transform theory) that if the crystal is “excited” by a very short, intense pulse of voltage (mathematically, a -function), then the observed signal will be an exponentially-decaying cosine wave. Excitation = (t ) Signal = + R( )e - it d , [18] where R() is the frequency response function. If R has the Lorentzian form above, then this Fourier integral leads to 17 Signal e t cos 0 t . [19] 3.6.4. Q-values and pulse lengths A second way of looking at the signal is in terms of an alternative definition for Q: Q 2 Energy stored Energy lost per cycle . [20] N.B. The denominator is actually shorthand for instantaneous rate of energy loss, expressed in units of energy per cycle Energy loss per second dE / dt 2 dE No. of cycles per second f 0 dt [21] Hence, Q becomes dE 0 E E E0e ( 0 / Q ) t dt Q . [22] The amount of energy stored goes down exponentially with time constant 0/Q. In the ideal case, all the energy loss of the transducer goes into the ultrasound beam. Hence we have I I 0 e ( 0 / Q ) t e i 0 t . [23] Finally, as we have seen, I = P2 / Z, so the pressure variations in the pulse of ultrasound have the form p p0e ( 0 / 2Q ) t ei 0 t . [24] This explains the way in which ultrasound pulses are drawn in the illustrations. It is convenient to define a “nominal” pulse length, the time taken for “almost all” of the power to have gone from the pulse. This choice is arbitrary, but it is often convenient to choose a length of time t = Q / 2 periods of the waveform. This leads to I I 0e( 0 / Q )( 2 / 0 ) / 2 I 0e 18 , [25] i.e., the intensity decays to around 4% of the original. 3.6.5. Resolution and Frequency (F184–185, C82–89, 104–112) It is the distance apart which two ultrasound objects must be to give distinct signals at the receiver. Axial (often called “longitudinal”) resolution is the quantity that is important for Amode scans and refers to resolution along the beam direction; Consider two reflectors a distance d apart, as in Fig. 14. When the pulse hits Interface 1 (a), some is reflected and some transmitted (b). The pulse then arrives at Interface 2, where, again, some is reflected and some transmitted (c). Interface 1 Interface 2 (a) Eventually, the two echoes arrive back at the transmitter, separated by twice the distance between the reflectors. 0 50 100 150 200 Incident Pulse When d is reduced to the point such that the echoes d are just resolvable, then (b) AR 200 150 100 50 Q 4 00 50 . [26] 100 150 200 (c) 200 (d) 200 150 100 50 Echo 1 150 100 50 200 0 150 100 50 0 0 50 10 0 15 0 20 0 Echo 2 0 200 150 100 50 0 0 5 0 1 0 0 1 5 0 2 0 0 0 50 10 0 15 0 20 0 2d 19 Figure 13: Reflection of a pulse of ultrasound from two closely separated interfaces The important points to take away are: Axial Resolution Wavelength Signal 200 150 100 200 50 + 150 0 Two peaks just resolved Demodulation 100 50 0 t AR Minimum resolvable distance Using a smaller wavelength (higher frequency) gives a better resolution. Figure 14: Criterion for resolution of two ultrasound reflectors 3.6.6. Attenuation and frequency In Sections 3.4.1 and 3.4.2, we noted that both scattering and absorption rise with frequency. This is bad news because the penetration depth of the ultrasound goes down. When choosing the frequency to use, we thus have to make a compromise between resolution and signal attenuation. 4. B-Mode (F69–96, C135–158, W348–351) 4.1. Introduction The commonest form of ultrasound imagingThe ultrasound equivalent of the radar pictures you see in those old war movies.A thin beam of ultrasound is scanned across the object and the strength of the returned echoes is displayed on the monitor.Notice that whilst in radar, full 360 coverage is required, in medical ultrasound, where only the body in front of the transducer is of interest, we look at a limited “pie-shaped” sector. A B-scan is “simply” an A-scan in which the ultrasound beam is moved and the results are displayed differently. 20 “B” means “brightness”: the ultrasound signal changes the brightness of a spot on an oscilloscope screen instead of the amplitude of the trace in A-mode. 4.2. What do we need to add to an A-scanner to turn it into a Bscanner? As soon as we try to turn the idea into a working system, we find a number of problems lurking! How do we display the data received?How do we make the beam sweep across the sample?What do our data mean?Fig. 16 is a block diagram of a generic B-scanner. Only three new items have been added: the co-ordinate generator, the video amplifier and the beam-steering device. 4.3. The Co-ordinate Generator This device is often also called the scan converter. It takes information about the instantaneous orientation of the beam and turns it into the co-ordinates of a line on the display monitor. 0 50 100 150 200 Reflecting Surfaces 0 50 100 150 200 Ultrasound beam Boundaries giving rise to echoes Image Formed Other orientations of ultrasound beam Figure 15: Principle of B-scanning 21 V t PRF generator New Pulse generator Beam steering device Protection circuit Probe Variable gain amplifier (TGC) TGC generator V(dB) t New Demodulator Compression and Video Amplifier Brightness Co-ordinate Generator (x,y) Display Figure 16: Block diagram of a B-mode scanner On simple systems, the CRT electron beam is physically scanned up and down the desired line (i.e., the co-ordinate generator acts as a variable voltage source to the ’scope x- and y- plates). On more modern systems, the co-ordinate generator gives the memory location in which signal information is stored. The data is then displayed on a monitor by an computer program. 4.4. Compression and Amplifier Even after passing through the TGC, the range of signals in the data is still large. 22 This is due to the range of reflector strengths in the body — see Fig. 5. The compression amplifier transforms the data by some rule Vout = f(Vin), which reduces the dynamic range of the data (i.e., compresses the scale). Typically, a 40–50 dB dynamic range for Iin (i.e., the ratio Iin max/Iin min 104–105) is transformed to an output dynamic range of 10–20 dB (10–100). Remember: take the square root of these values to get the corresponding voltage amplitude ranges. This allows low intensity echoes to be seen on the same display as high intensity ones, i.e., strongly reflecting organ boundaries and weakly reflecting internal structure can be seen on the same image. A video monitor can display only about 256 values simultaneously and the human eye can distinguish only about 32 different shades of grey. This means that: a huge amount of information is lost as in the case of the TGC;one should not normally interpret B-mode image intensities quantitatively. 4.5. The Beam-Steering Device This is what distinguishes the different types of scanner. There are various levels of distinction. The most basic is between static and realtime scanners. 4.5.1. Static B-Scanners The transducer is moved manually by the operator. Hinge Probe The probe slides backwards and forwards over the patient, changing its angle.The image is built up line by line. Each time, the co-ordinates 1, 2 and 3 tell the display where on the screen to show the results. See Fig. 17 Patient Fig. 17: Co-ordinate generator for static Bscanner The advantage of the system is that the operator can choose which bits of the picture to update most often and to tailor 23 Oil bath Oscillating mirror Ultrasound beam Rotating transducers Window Oscillating transducer Fixed transducer Figure 18: Various types of beam steering device for real-time scanners the scanning motion to view the feature of interest from several different directions It is also very cheap. However ... The scans take several seconds to build up and form a complete picture. This is a problem if the object in question moves in the meantime. Static B-scanners are not suitable for imaging, for example, a beating heart. 4.5.2. Real-Time Mechanical Scanners (F83–84, C144–146) “Real time” scanners acquire anything from a few frames (images) per second up to several hundred. They are ideal for imaging motion. In a motorised scanner, the transducer is moved mechanically by a motor. Because of the difficulties of maintaining contact between the skin and a moving transducer, a larger probe is used, which contains the transducer “suspended” in a bath of oil, with a window to allow the pulses to leave. There are several different designs, as shown in Fig. 19. In all cases, the final device will depend on obvious mechanical engineering questions like: How do you make a probe rock backwards and forwards very fast? Can you make it do so uniformly? How do you get leads to three transducers on a ring without everything getting tangled up when they rotate? 24 The major disadvantage of this type of device is that mechanical systems have an inherent speed limit. The advantage is that there is no complicated (and expensive) electronics. 4.5.3. Electronic Steering — Transducer Arrays (F89–94, C146–158, W340–343) We shall not go into any detail until later, but the basic principle is that a number of very small transducer are placed into a line and are then fired separately. By “firing” (i.e., sending out a pulse from) the transducers at different times, one can make composite wavefronts (Huygens Principle again!) which mimic that given out from one of the moving transducers above. Electronic beam steering is potentially much faster than mechanical steering and also has the advantage that the order of sampling of the different lines is much more flexible. 5. The Theory of Diffraction 5.1. Making Huygens’ Wavelets Quantitative We are all used to the simple concept of Huygen’s wavelets, as shown pictorially in Section 3.4.3. But how can we use this mathematically and without having to draw all those fiddly diagrams? Fresnel proposed that we could represent all the points on the wavefront by individual sources of spherical waves.In principle, each spherical wave would give rise to a contribution to pressure p e i (t kr ) r where r is the distance from the point on the wavefront (i.e., the origin of the secondary wavelet) to the point where we are making a measurement. 25 [27] y x O X R Y r Q P Figure 20: Notation for diffraction theory In practice, there are a number of complications and proportionality factors that we will not consider. Consider instead , which we may regard as being the same as p, except for a few constant factors and 2, which we will take to be the same as I. Consider the situation shown in Fig. 20. Waves are incident from the left ( x, y ) ei (t kx ) [28] These hit an obstacle at x=0 and we look at the Huygen’s wavelet at a typical point Q with height y compared with the origin O. We are interested in evaluating the contribution of this wavelet at point P. We need to describe the obstacle. This is done by an aperture function, a(y). In the simplest case, a(y) is 1 where the wave can get through and 0 where it is blocked. In more complex examples, we can vary the amplitude of a(y) to represent partial transmission or introduce an arbitrary phase. d (at P) d ( X , Y ) a( y )dy ei (t kr ) r , [29] where r = (X2 + (Yy)2)1/2 is the distance from the point Q to the point P. Overall value of is the result of summing all the individual contributions d, i.e., doing an integration. 26 Y-y Difficult to go further without approximations … Replace r as follows: r X R 1 2 Y 2 2 yY 2 yY R2 1 / 2 where R is the distance from O to P, i.e., (X2 + Y2)1/2. Now use a Taylor expansion yY r R 1 2 R R yY R R y sin where is the angle between the straight-ahead direction and the point P, as measured w.r.t O. Hence Eq. [29] becomes d ( X , Y ) a ( y ) dy eik ( R y sin ) R y sin We then argue that, whilst the extra phase given by the ysin term is important, the extra amplitude is not (for y<<R) Finally, the expression for the total signal at P is given by ei (t kR ) ( X ,Y ) R a( y ) e iky sin dy [30] Fraunhofer integral This important equation tells us that the waveform depends on the angle rather than on the particular values of X and Y. So we make a further notational change and rewrite as 27 (q) a ( y ) eiqy dy [31] Here, I have been very cavalier and “got rid of” the term before the integral, because it’s not very interesting. The key bit is the integral. Introducing q = k sin is a minor but important change, the purpose of which is to make the equation look like a Fourier transform. This is the form of the Fraunhofer integral that we will use in practice to tell us about our diffraction pattern. 5.2. The plane piston transducer Simplest type of transducer — a rectangular block. We will look at the problem in 1-D but extension to 2- and 3-D is straightforward. 1 - b/2 y b / 2 Aperture function a ( y ) 0 otherwise Notice that this is just the same as for a single slit of width b. The Fraunhofer integral becomes (q) b/2 1 eiqy dy b / 2 b/2 eiqy iq b / 2 b sinc b2 q Notice that what we end up with is a function of q = k sin This can be displayed in one of several ways — see Fig. 21. 28 (a) 1 2 1 (c) 0.8 0.6 0.6 0.4 0.4 0.2 sin 0.2 0 -0.2 0 -1 -0.5 0 0.5 sin 0.8 0 0.2 0.4 0.6 0.8 1 cos 1 -0.4 (b) -0.6 -0.8 -1 Y = R sin Figure 21: Possible methods of display for the function . (a) is a plot of intensity ( v. sin . (b) is the equivalent pattern of fringes as might be observed in an optics experiment. (c) is a typical transducer beam profile. (Note that in order to visualise the sidelobes (which are very small, I have plotted not 2. These diagrams were simulated for the case of b = 5.3. Two very small transducers Young’s Slits 1 - b y a and a y b Aperture function a ( y ) 0 otherwise [32] Since the Fourier transform is linear, we can regard this as the sum of two terms: (q) b a 1 e iqy dy b 1 e iqy dy a b sinc b2 q a sinc a2 q [33] Note that when the slits are very thin, we can formulate the problem in a different way: a ( y ) ( y b) ( y b) [34] which leads to the Fraunhofer integral (q) ( y b) e iqy dy ( y b) e iqy dy Remembering that the integral of the delta function has the effect of sampling the accompanying function at the location of the delta function, we get 29 (q) e iqb e iqy 2 cos qb [35] This result is very famous and describes the pattern of fringes that is seen in the classic optics experiment Young’s slits. 5.4. The linear array transducer A key question arises: Can we generalise the result above for more than two transducers? This is important for work using linear arrays. Consider the array in the Fig. 22. Note first that this is a 2-D problem. Diffraction both in the X- and Y-directions. It turns out that the result is simply the product of one function for Y, which we have already calculated as h sinc qh/2, and a term corresponding to the ax(x) aperture function. In class we will show mathematically how we can use the Convolution Theorem to find a general result that will apply to linear arrays with any number of elements. The result is the beam pattern shown in Fig. 23. It possesses an important feature — the existence of major beam sidelobes. These will be considered in more detail later. h L Y d X b Figure 22: Schematic diagram of a linear array transducer 30 1.2 (a) (b) sin 1 1 0.8 0.5 0.6 0.4 0 0.2 0 0.2 0.4 0.6 0.8 0 -1 -0.5 0 -0.2 0.5 1 sin -0.5 -0.4 -1 Figure 23: Beam pattern of the linear array transducer presented in two different forms. Here the results have been simulated for the case where b = and d = 3. 5.5. Electronic beam steering re-visited The quantitative theory of diffraction developed above can explain how to steer an ultrasound beam. The diffraction pattern above is what we get if we fire the elements all at the same time; beam steering involves firing the elements at different times. If the elements are made to fire at intervals t, such that t d sin c , [36] then the wavecrests will form a line at angle . This corresponds to a plane wavefront moving in the direction , i.e., we have steered the beam electronically. Fig. shows this pictorially, but we can also calculate the results using the Fraunhofer diffraction theory. Each time delay t corresponds to a difference in the phase at which the Huygens wavelets are emitted. 2 j t (ck t ) j T , where j is the number of the transducer. 31 [37] 1 cos Phase shifter introduces time delay t in signal Transducer t = 0 For this transducer, t = (6d sin /c d t = (d sin /c t = (2d sin /c t 0 0 0 50 10 0 50 15 0 0 50 10 0 15 0 0 50 10 0 10 0 15 0 20 0 50 15 0 Effective plane wavefront 6d 10 0 15 0 20 0 20 0 20 0 20 0 Figure 24: The principle of electronic beam steering The expression for ax(x) now becomes a x ( x) ei j j ( ck t ) ( x jd ) hb ( x) hL ( x) [38] Showing how this expression leads to beam steering is one of the problems on the examples sheet. Electronic beam steering has the following advantages: (i) The beam motion can be faster than by mechanical steering. (ii) The order of sampling does not have to be a single sweep backwards and forwards. Hence the screen refresh rate can be more uniform, as opposed to a sort of “windscreen wiper” effect. In addition, we avoid confusion between outgoing and incoming signals from adjacent lines. (iii) It is easy to arrange to update some lines more frequently than others, to concentrate on, for example, a beating heart. Set against these, there are two main disadvantages: 32 Motion of transducer Image Profile across image Small transducer Signal strength Ideal, thin US beam Object a a a x Blurred image Signal strength Real US beam width b b a a+b x a+b Figure 25: Blurring of an image due to a wide ultrasound beam (i) The processing electronics are more complicated. This is becoming less of a problem as electronics prices fall. (ii) The side-lobes associated with the linear array lead to problems. There is ultrasound intensity where we don’t want it. If this bounces of a strong reflector and gets back to the receiver, we get “ghost” images, as described in the next section. 5.6. Lateral Resolution, Diffraction and Focussing Lateral resolution is the smallest distance that we can distinguish perpendicular to the beam and in the image plane. Obtaining the best possibile resolution depends on getting the smallest possible beam diameter at the point of reflection. Diffraction effects mean that we can’t get the beam as sharp as we would like. If we start off with a narrow beam, it will spread out. 33 Far-field (“Fraunhofer”) region Transition zone d2/4 d Diverging beam(sin = 1.2/d), but smooth variation of intensity within the beam. Near-field (“Fresnel”) region Tightly collimated beam, but internally, a complicated structure containing lots of maxima and minima of signal intensity - not much good for imaging Figure 26: The three regions in the beam pattern of a circular transducer A mathematical treatment shows us what the beam from a simple transducer will look like. We can divide it into three regions: near-field, transition zone and farfield, as shown in Fig. Only the far-field region is suitable for ultrasound imaging. Consider the intensity variation of the beam along its axis. A mathematical treatment of the diffraction (see 2US) shows that at a given time t, I( x) I0[ cos(t kx) cos(t k a2 x 2 ) ]2 . [39] The important feature is the “envelope” or time-average of this function. You do not need to know this formula, nor the exact details of the pattern shown in Fig. 27, but you should be aware of the consequences: (i) In the near field, the rapid oscillations of the signal intensity make it useless for imaging.In the far field, the intensity varies smoothly.(iii) The closest point for imaging is approximately 34 Fresnel Transition Fraunhofer Monotonically decreasing ultrasound intensity Complicated intensity patterns d2/4 0 Figure 27: Simulation of signal intensity along beam axis. This is a “snapshot” at time t of Eq. [39]. In the Fraunhofer zone, we can regard the function as a cos kx function with a slowly decreasing envelope. dmin a2 / D2 / 4 . [40] (iv) Since we use the Fraunhofer zone for imaging, with a diverging beam, the lateral resolution gets worse with increasing depth. A limited improvement can be achieved by using ultrasound lenses to focus the beam. However, the mathematics of diffusion say that you can only go so far: (i) No matter how good the lens is, there is a minimum beam diameter. In the focal plane of the lens, the beam size for a circular transducer is d 2.4 flens , D [41] where flens is the focal length of the lens and D the diameter of the transducer. This is typically of the order 0.2mm. (ii) The aim of focusing is to get a focal spot with size d<D. Hence, 35 2.4 flens D D flens D2 . 2.4 [42] I.e., focusing does not work at all for distances into tissue greater than a certain value. (iii) The tighter you focus the beam, the less long it stays focused for — see Fig. 29. We have to make a compromise between width of beam at focus and depth of field. Focal Plane Ultrasound lens Diverging Beam D Transducer flens Focal Spot (diameter, d) Figure 28: Use of an ultrasound lens to focus the beam 36 (a) (b) Figure 29: (a) Strongly focused and (b) weakly focused ultrasound beam 5.7. Artifacts … or … Why do ultrasound images look so awful? (F179–189) It takes an expert to “read” most ultrasound images. Why is this? There are a number of artifacts to which ultrasound images are prone. 5.7.1. Misregistration A bright spot in the image appears in the “wrong” place. I.e., it does not correspond to the true position of the reflector. There are several reasons for this: (i) Refraction effects — see Section 3.3 37 Transducer Ultrasound Image Main beam Weak, angled beam (“side-lobe”) Weakly reflecting object Strongly reflecting object Figure 30: “Ghost” image caused by transducer side-lobe (ii) Velocity differences Since depth in ultrasound images is given purely by the time between the transmission of the pulse and the reception of the echo, then if any part of the medium between the transducer and the reflector has a non-standard velocity, then the wrong depth will be recorded. The bright spot will appear at the wrong place in the image. (iii) Side Lobes In Sec. 5.4, we noted that the transducer beam had a central lobe, plus sidelobes of reduced intensity at other angles. If one of these beams reflects off something, then an echo will be received from one angle when the detector “thinks” it’s pointing at a different angle — see Fig. 30. (iv) Multiple Reflections It is possible for the ultrasound beam to bounce backwards and forwards several times between the transducer and shallow reflectors. At each stage, the transducer will register an echo. This gives rise to a number of evenly spaced spots on the final image. 38 Figure 31: Typical abdominal ultrasound image. The arrow is pointing to an example of acoustic shadowing, but this is not obvious to a non-expert. (Image source: Brigham Women’s Hospital, Harvard) 5.7.2. Shadowing and Streaking This is a consequence of the TGC unit — see Sec. 3.5.3 Suppose we have a region of very low attenuation. This lets through more ultrasound than the TGC is expecting. When this reflects off the objects behind, we get a streak with higher intensity than normal. Suppose we have a region of higher than normal attenuation. This lets through less ultrasound than expected the signal from objects behind it will be less than expected. This gives rise to a shadow. Both these artifacts can be useful diagnostically. 5.7.3. Speckle Possibly the worst artifact.Caused by having small reflectors (not the infinite planes used to derive the equations of reflection), which are closely spaced.An interference effect. The reflected pulses interfere with each other in a random way — see Fig. 31.BlurringThis is caused by the limited resolution of the apparatus — see Sec. 3.6.5.Separate blurring occurs in the axial, lateral and slice directions. 39 5.7.5. If ultrasound images are so bad, why use them? (i) As an imaging method ultrasound is relatively cheap.(ii) It’s quick — continuous, real-time images of, say, a beating foetal heart, which is difficult by any other technique. (iii) It’s portable — you can take the machine right to the patient’s bedside.It’s very patient friendly (v) The human eye is very good at making sense of noisy data and recognising the patterns which correspond to the features of interest. Trained radiologists can obtain very large quantities of information. (vi) A general rule of imaging: pretty pictures are to impress the punters. If you can get all the information you want easily from the given data, then you don’t need anything more sophisticated. E.g., obstetric ultrasound: the recognised measurements are diameter of foetal skull and length of femur. These can be made to within a few mm with cheap, standard equipment. 6. Doppler Ultrasound (F117–164, C186–212, W351–358) 6.1. Introduction Measurement of blood flow is the second big use (after obstetrics).This is done using the Doppler effect.When ultrasound waves strike a moving target (e.g., a red blood cell), the reflected wave is shifted in frequency.This frequency shift is proportional to the velocity of the target. Hence we can measure the speed of blood flow. 6.2. The Doppler Equations Fig. 32 shows the notation we will be using. Because the blood is moving w.r.t. the transducer, the effective velocity of sound relative to the red blood cells (RBC) is vrel,inc c component of v along direction of beam c v cos inc So, wavecrests arrive at the RBC with a frequency 40 [43] fRBC = vrel, inc (normal v f with changed speed) inc c v cos inc inc [44] These wavecrests are scattered off the RBC and are re-radiated back to a second transducer. Clearly, since the source is moving away from the receiver, these waves are “stretched out and so scat is greater than inc. Since the frequency of re-radiation of wavecrests by the RBC is fRBC, then in a time t, fRBC t wavecrests have been emitted. During this time t, the first wavecrest will have travelled a distance ct in the laboratory frame of reference, but ct + v cos scatt relative to the RBC, which is receding from it. Thus, there must be fRBC t waves contained in a distance ct + v cos scatt. The reflected ultrasound beam has wavelength scat ct v cos scat t fRBCt (c v cos scat ) 41 c v cos inc inc [45] Figure 32: Statement of the problem in Doppler ultrasound. A moving boundary (RBC) scatters an incident ultrasound beam and the reflected beam is shifted in frequency. The final result can be written either in terms of the reflected wavelength, or, more usefully, in terms of the frequency of the reflected wave in the laboratory frame of reference. scat c v cos scat inc c v cos inc or fscat c v cos inc finc c v cos scat [46] We often make the simplifying assumption that v<<c, in which case, the frequency difference between the two beams is v f finc fscat (cos inc cos scat ) finc c [47] 6.3. How do we put all this into practice? Fig. 33 shows the block diagram of a Doppler ultrasound scanner. This most primitive type of Doppler machine is called a continuous wave (CW) scanner. It does not emit pulses like an A or B mode scanner. 42 Figure 33: Block diagram of a CW Doppler ultrasound machine What is reflected back is a continuous signal which contains some ultrasound reflected by moving spins and some reflected back by stationary spins. The extra components of the Doppler scanner are listed below. 6.3.1. Second transducer for reception The Doppler probe in the diagram above consists of two transducers. The first transducer is transmitting ultrasound continuously. There is no “spare” time to receive the signal bounced back. The second transducer is normally very close to the first and offset at a very slight angle. incref and so Eq. [47] becomes f 2v cos inc finc c 43 [48] (a) Spectral Amplitude A f0 = 2(v/c) cos inc v f0 Demodulated frequency f A (b) fpeak = 2(vpeak/c) cos inc vpeak fpeak (c) Rapid flow past obstruction f A Stationary object peak Slow flow downstream of obstruction Stenosis (blockage) “Negative” 0 flow in eddy Reverse flow in eddies f Figure 34: Three different flow patterns with their corresponding Doppler frequency profiles, as given by a spedtrum analyser: (a) plug flow, only a single velocity present; (b) Poiseuille flow with quadratic velocity distribution — common in vessels with slow flow; (c) complex flow pattern around an obstruction in the vessel. (a) and (b) represent “idealised” patterns; in practice, one would not obtain such “clean” patterns. 6.3.2. Doppler demodulator This works in exactly the same way as described in Sec. 3.5.4. Note that the frequency which we subtract from the signal is finc, which is used as a reference. Thus, in the block diagram, the output from the oscillator goes both to the transducer and the demodulator. 6.3.3. Frequency meter / spectrum analyser After the demodulation process, we have an oscillating signal of form cos (2ft). We still have to extract the value of f. On primitive machines, radiologists just listened to the signal, since f is in the audio frequency range. 44 f v Range of flow velocities present at time t0 t0 t Figure 35: “Real time” graph of flow velocities in a vessel. Slightly more sophisticated is a frequency meter, which counts the number of zerocrossings of the function, divides by 2 and displays the result as a frequency. This leads to problems with multiple frequency components and noisy data. A real-time spectrum analyser is the ideal solution. It gives a graph of all the different frequency components. 6.3.4. Real-time display Displays such as Fig. 34 show complete information for a “snapshot” view. However, it is often more useful to show more limited information, but in real time. Modern Doppler machines can acquire the information very rapidly, so we often plot a graph with time along the x-axis and velocity range up the y-axis. By analysing such plots we can find out about the flow properties of blood vessels and diagnose problems. 6.3.5. Problems with CW Doppler (i) Signal comes from the whole sensitive region — see Fig. 33 — we have no way of deciding where the flowing material is. What happens if there are two vessels with different velocities?We have to estimate the value of [48], so the derived velocities are not certain.We can only find the component of velocity along the beam direction. If the scatterer is moving perpendicular to the beam, there is no Doppler shift.Duplex Scanners The ideal solution to problems (i) and (ii) above is to combine a Doppler ultrasound scanner with a B-mode imager. 45 To do this, we need to select where we acquire the Doppler information from. This process is called range gating and works as follows: 1. A pulse of ultrasound is emitted at angle 2. We wait for echoes to come back, but turn the receiver on only between times The frequency shift of the signal is analysed as before (although there are a number The whole procedure is repeated for region.The velocity information is often colour-coded and superimposed on a normal greyscale B-scan. 6.5. Limitations of Doppler Ultrasound Although Doppler ultrasound has revolutionised many aspects of diagnosis, it is important to bear in mind that there are a number of limitations: (i) We can measure only the component of flow along the direction of the ultrasound beam.We need to be sure of the angle between the beam and the vessel of interest. Blood vessels are rarely straight and this can lead to problems in assigning the velocity correctly.The blood cells from which the ultrasound is scattered are not all moving at a uniform velocity, so a spread in values will be obtained. Flow is often turbulent.In practice, although it is possible to undertake much more sophisticated forms of analysis, clinicians use Doppler ultrasound in a very crude way. For example, a typical decision might be: flow reduction < 50%, don’t operate; flow reduction > 50%, think about operating. 7. Bio-effects of Ultrasound (F201–212, C213–218, W375–377) 7.1. Introduction Sufficiently strong intensities of ultrasound have the potential to affect living tissue.In some cases, these effects are beneficial and can be used for therapy.However, in diagnostic imaging, we wish to keep exposure as low as possible.Heating 46 Tissue mass m A Beam propagation x+x x Figure 36: Notation for calculating temperature rise in a tissue caused by absorption of ultrasound energy In Sec. 3.4.2, we saw that an ultrasound beam is attenuated as it passes through tissue and that the beam’s acoustic energy is changed into heat. How does this change the temperature of the tissue? We have IRMS(x) = IRMS(x) exp(2x), or, in differential form, I RMS ( x) 2I RMS ( x) x . [49] IRMS(x) is the power loss to the beam, so, by conservation of energy, it must be the power gain by the tissue. Remember: is the attenuation coefficient for pressure, I0(x) is the peak power flux through a unit area (measured in W/m2). It can be shown that the average power is half this. The energy gained by the shaded tissue element in time t is Q 2 AIRMS ( x) x t. [50] Assume that heat conduction by the tissue is quite poor and that we look at the temperature rise only for the first few seconds after the beam is switched on [51] where c is the specific heat capacity and m is the mass of the volume element, i.e., (Ax). This gives 47 T 2 AIRMS ( x ) x t 2IRMS ( x ) I t 0 t, A x [52] or, alternatively, the initial rate of temperature rise is dT 2IRMS . dt [53] In practice, selective heating of tissue is used as a method of treating cancer. The ultrasound beam is focused so that where the tissue is not tumorous, IRMS(x) is small and the heating rate is low, but at the tumour a large amount of power is concentrated. In fact, the assumption that we made above is a gross over-simplification. The conduction by tissue is not negligible and in practice, one needs to use a heat diffusion equation. In addition, the body has its own thermo-regulatory system, which tries to keep the body at its normal temperature. Thus, flowing blood can carry away excess heat quite efficiently. The organs most susceptible to clinical hyperthermia are those with a poor blood supply. With a less focused beam, the heating effect is lower, but more widespread. This is often used to apply localised deep heat to damaged muscles in physiotherapy. 7.3. Radiation Pressure You will find in many branches of physics that a flow of energy via a wave motion brings with it a flux of momentum. The normal relationship is E Mc [54] where M is the momentum “carried” and c is the speed of the radiation. For ultrasound, the energy carried per second is AIRMS(x). (Remember: A is the cross-sectional area of the beam and I0(x) is the peak power flux per unit area — the time-average is half this.) The momentum carried is M AIRMS ( x ) t. c 48 [55] Newton’s 2nd Law tells us that Force = Rate of change of momentum and the definition of pressure is Force/Unit area. Hence Prad I RMS ( x) . c [56] If, instead of being absorbed, the ultrasound is reflected, then the momentum changes from M to -M (i.e., the total change is 2M) and so the pressure is doubled. This force is extremely small, but may affect delicate structures in the body. 7.4. Acoustic Streaming Putting together the arguments of the previous two sections, it is clear that each element x absorbs a power aIRMS(x)x per unit area, which is equivalent to a momentum of IRMS(x) x / c per second. Hence, there is a force F = IRMS(x)x / c acting on each element xof the sample. If that sample is a liquid, it moves. If the liquid is in an enclosed vessel, we get a characteristic type of motion called acoustic streaming. Streamline Fluid motion Transducer Figure 37: Acoustic streaming in an enclosed container A delicate structure in such a stream may be affected both by the direct force of the ultrasound, or by viscous drag from the liquid, or even Bernoulli forces. 7.5. Cavitation Suppose that there are tiny bubbles in the sample, with diameters <<l. When an ultrasound wavefront (pressure maximum) coincides with the position of the bubble, the bubble is squashed.Half a cycle later, there will be a rarefaction (pressure minimum) and the excess pressure in the bubble causes it to expand.So, 49 as the peaks and troughs pass by, the bubble oscillates in size. Overall, the bubble may grow by a process known as rectified diffusion. During the accoustic pressure cycle, gas alternately diffuses in and out of the bubble. In the rarefaction phase, the surface area of the bubble is larger and more gas diffuses in during this part of the cycle than diffuses out during the compression phase. Hence the volume of the bubble increases. There are two types of cavitation, stable and unstable. Both types have the potential to cause damage to tissue. In stable cavitation, the bubbles persist for a long time (i.e., a relatively large number of acoustic cycles) and can grow quite large. They cause micro-movements of the surrounding liquid and tissues. This may be harmful. In unstable cavitation, the bubbles burst violently and this can cause important changes at a cellular level. The bursting is associated with high local temperatures, emission of light, chemical changes and the formation of free radicals, which can be very destructive. (a) Bubble Crest of ultrasound wave c Low external pressure (b) High external pressure Figure 38: Mechanism of cavitation. (a) In a low external pressure (rarefaction), the vapour pressure inside the bubble causes it to expand. (b) When the external pressure is high, the bubble contacts. As the ultrasound wave passes, the bubble oscillates in size. 50 7.6. Therapeutic Ultrasound 7.6.1. Physiotherapy Several of the effects described above are believed to play a role in the treatment of muscle damage in physiotherapy. Although not all of the mechanisms have been fully explained, it is known that ultrasound is beneficial for: increasing blood flow and nutrient supply to the damaged muscle (by a factor of up to 2 or 3);breaking up unwanted/damaged/fibrosed parts of the muscle;aiding transport of chemicals across cell walls; relieving pain — this is not well understood;increasing tendon flexibility;reducing muscle “spasm” — cramp is an extreme form of muscle spasm.Average power (IRMS) used in physiotherapy ranges from about 0.125 to 3 W/cm2.Surgical Uses This makes use of two main properties: The heating effect is used in the treatment of cancer by controlled clinical hyperthermia. Essentially, the tumour is “cooked” to kill it. The mechanical vibration and cavitation effects (also used in “sonicator” cleaners/stirrers in chemistry) can dislodge unwanted tissue or particles. It is often used to fragment kidney stones (lithogtripsy) and sometimes cataracts as well. 7.7. And finally ... is it safe? One of the reasons why diagnostic ultrasound is used in pregnancy is that it is claimed to be completely safe. However, as we have seen above, it is used at higher powers to kill cancerous tissue deliberately. So what power is “safe”? We can give the following rough guidelines: IRMS << 0.125 W/cm2 0.125 < IRMS < 3 W/cm2 IRMS >> 3 W/cm2 diagnostic — “completely safe”, “does not change tissue” physiotherapy — “safe”, “does not harm tissue” surgery — damages/kills/shatters tissue in a localised area 51 As with any other technique, it is impossible to prove that ultrasound (even diagnostic ultrasound) is completely safe. We can only demonstrate the scarcity of recorded cases demonstrating that it is dangerous. 52