- Email: [email protected]
Accuracy of colour Doppler ultrasound velocity measurements in small vessels
Accuracy of colour Doppler ultrasound measurements in small vessels C.R. Deane, F. Forsberg,
velocity
N. Thomas and V.C. Roberts
Department of Medical Physics and Medical Engineering, King’s College School of Medicine and Dentistry, Dulwich Hospital, East Dulwich Grove, London SE22 8PT, UK
ABSTRACT ColourDoppler ultrasound offers thepossibility of imaging small vessels not visible by B-mode alone. The colour Doppler image of velocities allows the course of small vessels to be imaged in the X-Yplane of the scan provided the Doppler s E 1so suI?’ nent magnitude. This permits alignments of the Dapple r cursor, allowing angle correction to f;qe uenqhlfi’f provide true velocity measurements from the Doppler shiJi obtained. Before attempting to make velocity measurements, however, it is essential to be aware of the possible error in the Z plane caused by the thickness of the Doppler sample volume. To quantzfi this source of error, hydrophone and flow-rig measurements were peformed on an Acuson 128 colour Doppler scanner with both 5MHz linear-array and 3.5MHz phased-array transducers. Measurements of the transmittedpulses using a point hydrophone showed that both probes employ approximately 3.5MHz Dopplerpulses (in both colour andpulsed Doppler modes). The two transducers have the same axial resolution. In colour Doppler mode the axial length of the sample volume increases automatically with depth by up to 0.5mm. Measurements of colour and pulsed Doppler signal strength were obtained in a controlledJl ow rig. Both transducers produced accurate colourJow images of the phantom at their optimum depths; flow velocity errors due to Z-plane thickness are < 5%. There was, however, substantial error outside these optimum conditions (up to 20%). Keywords: Colour Doppler
ultrasound,
hydrophone
measurements,
INTRODUCTION Colour Dop ler ultrasound (CDU) scanners have been availab Pe for vascular investigations for over 2 years. The use of colour Doppler images to scan an area for Doppler frequency shifts offers advantages when compared with conventional duplex Doppler scanning, where only a specific volume or ‘gate’ is analysed at any time: it provides rapid assessment as to whether a structure under investigation is vascularized or not, highlights areas of high or abnormal flow; and can assist pulsed wave Doppler (PW) investigations by allowing rapid positioning of the PW sample volume. CDU also offers potential quantitative information in showing the course of small vessels not visible in Bmode images alone. This permits PW angle correction to allow measurement of arterial and venous velocities by PW using the classic Doppler equation: fi,=
2v cost3 ---fc
(1)
cb
where ft, is the Doppler frequency shift measured, u is the velocity of the blood, 8 is the angle to be corrected between the ultrasound beam and the vessel, ct is the velocity of sound in tissue and fcis the transmitted ultrasound frequency. The accuracy of the angular correction is, however, largely dependent Correspondence
to: CR.
tests, Acuson
128
on the resolution of the colour Doppler image. This is ap arent in the plane on the screen (the X-Y plane 7 . What is not readily appreciated is the ambiguity in the Z plane caused by the beam and sampling width of the scanner when in colour Doppler mode, as illustrated in Figure 1. In this paper we present work undertaken on an Acuson 128 colour Doppler scanner. The aim was to investigate the way in which the colour Doppler scanning is accomplished and to attem t to quantify the errors obtained when using a co Pour Doppler image to provide the pulsed Doppler beam-vessel angle correction, 0. The work falls into two parts:
\ t!L
1. Using a hydrophone,
examination
of the emitted
Beam width
-I
iy
Vessel
\ \J/’ Colour Doppler
sample volume
Figure 1 2-D Doppler
Deane
0 1991 Butterworth-Heinemann 0141-:i42.5/91/030249-06
flow-rig
image representing
a 3-D blood vessel
for BES J. Biomed.
Eng. 1991, Vol. 13, May
249
Colour Doppler ultrasound accuracy in small vessek CR. Deane et al.
signals in B-mode, pulsed Doppler and colour Doppler modes. 2. Assessment of the accuracy of colour and pulsed Doppler measurements made on a controlled flow through a latex tube of known dimensions.
Gould 4050
w
oscilloscope
Absorbing foam
The Acuson 128 (Acuson, Mountain View, CA, USA) is a commercial ultrasound scanner with both pulsed Doppler and colour Doppler facilities. The transducers used during these studies were: Acuson I538 linear-array transducer, nominally 5 MHz with an aperture of 38 mm and a maximum depth of image of 80mm. S328 phased-array ‘sector’ transducer, 0 Acuson nominally 3.5 MHz with a maximum depth of image of 240mm.
IEEE 488
0
Very little technical information is available to users of the scanner; the information provided is primarily intended for operating the machine”L. The CDU investigation area may be altered; in the case of the I538 probe, from full frame to an area of 10 X 10mm. When the CDU area is reduced the frame rate is increased slightly and the colour pixel size becomes smaller. Similar changes occur with the S328 probe.
HYDROPHONE MEASUREMENT SOFTHE TRANSMITTED PULSE The resolution of the colour Doppler image is constrained by the size of the sample volume. For pulsed Doppler (including colour Doppler) the axial resolution of the sample volume, L,, is dictated by the duration of the pulse (number of periods within the pulse) and the filter characteristics of the receiver”T4 according to: =
L
experimental
set-up
for
the
pulses for the colour Doppler signal linear-array transducer are shown in Figure 3a-c for three different depths (0, 30 and 70 mm respectively). The Doppler frequency emitted is 3.5 MHz in spite of the B-mode, as specified for the transducer, operating at 5MHz. The measured values of both amplitude and pulse length are given in Table 7. The axial length of the sample volume due solely to the duration of the emitted pulse, L,,, was calculated according to equation (2). In practice, the axial length is also dependent on T,; however, absolute changes to Lxt contribute direct1 to L,. The observed increase in (peak-to-peak r amplitude from 110 to 200 mV is implemented to counteract the effect of attenuation on the signal-to-noise ratio (SNR). H owever, the increase in the number of periods emitted (from 4 to 7) was quite unexpected. The equivalent change in L,, is 50% (c$ Table 7). The result is that the axial resolution is reduced with depth. The same set of measurements have been performed using the 3.5 MHz phased-array transducer (S328). In this case the B- and colour Doppler-modes were found to employ the same centre frequency of 3.5 MHz. However, a similar reduction in colour Doppler axial resolution for an increase in depth of on the 5MHz
where Tp is the duration of the emitted pulse, T, is the duration of the receiver filter impulse response and cb is the average veloci of sound in blood (approximately 1570m s-‘) (re 9erence 5). It was not possible to investigate directly the characteristics of the receiver filter. However, to assess the sources of error arising from the contribution of pulse duration to the axial length of the sample volume, we measured T,, for both transducers in all modes at a range of depths. Methods
The transmitted pulses were measured in a water bath containing distilled water. The transducer in question and a 1 mm PZT point hydrohone (Mao-8X, A0 Williams, Malmesbu UK) were mounted on a fixture at a variable 7 ittance apart. After a manual alignment to secure maximum output neither the transducer nor the hydrophone was moved. A block diagram of the experimental set-up is given in
250
of the
phone was sampled directly using a Gould 4050 digital storage oscilloscope, which has a resolution of 8 bits and a sampling frequency of 100MHz. The sampled signal was transferred to a Compaq 386 personal computer, via an IEEE-488 interface, for processing and dis lay. from the The two trans cfucers were controlled Acuson 128 colour Doppler scanner. To obtain CDU pulses at a range of depths, the area of colour Doppler display on the B-mode image was reduced to a minimum. On an Acuson 128 the smallest area possible in colour Doppler studies is a square of 1 x 1 cm. All depths given in the following refer to the top of this square area. An absorbent foam was placed in the water bath to prevent reflections from the side of the bath interfering with the original waveform.
The recorded
2
Figure 2. The high frequency
Block diagram measurements
Results
dTp+ 7;)
x
Figure 2 hydrophone
signal from the point hydro-
J. Biomed. Eng. 1991, Vol. 13, May
CohT Doppler ultrasound accuracy in small vessels: C.R. Deane et al. Table 1
0.151
O.lj
7
0.051
d f g
Measurements on the Acuson L53H in colour Doppler mode
Depth
f
Amplitude
(mm) 0 30 70
(MHz) 3.5 3.5 3.5
(mv)
(?s,
[Irn)
100
I .32 I .70
1.0 1.3 I .5
200 190
I.95
01 Table 2
5
-0.051
Measurements
on the Acuson S328 in colour Doppler mode
Depth
f
Amplitude
(mm)
(MHz)
(mV)
2s)
I (zn)
6: 180
3.2 3.2
200 8.5 200
1.84 I .43 1.93
I.1 1.4 1.5
-O.l!
-0.15’ 0
1
2
3
4.5
i
a 1.5 to 30 mm, the number of emitted periods changes from 8 to 15, as illustrated in Figure 4u and b. These measurements, taken on the Acuson L538 transducer, are equivalent to a change in Lxt from 1.7 to 3.4mm, and demonstrate the direct relationship between pulse duration and sample volume axial length. A similar behaviour was observed for the 3.5 MHz At sample gate settings phased-array transducer. greater than this (up to 30mm), there is no further increase in the number of transmitted periods. It is
Time (x 10m6 s)
a 0.15’ 0.1 p
0.05t
0” 3 C
0’
g -0.05-
-0.q
-0.15L 0
7
8
Time (~10~~ s)
b
a 0.151
-0.02 t -0.03/
0.11
-0.04b 2
0.05
$ 2 “a E
Time (~10~~ s)
a
o;_
0.04
4-o.05t I
0.03-
-0.1; -0.15 i 0
C
f
s
Acuson
L.538
Time (x 10v6 s)
Figure 3 Emitted colour Doppler pulse of the transducer focused at: a, Omm; b, 30mm; c, 70 mm
the region of imaging was measured as shown in Table 2. Since the two transducers reduce almost identical Doppler pulses (compare $ ables 7 and 2) the S328 pulse is not shown. the transducers’ corresponding For comparison, PW pulses were studied as well. In PW mode, the sample gate is under direct user control; the axial length may be varied from 1.5 to 30 mm. In the range
-0.031 -0.04&
b
Lo
1
2
._
1~
~~~
3
4
5
6
7
8
Time (~10~~ s)
Figure 4 PW pulse of the L538 recorded at 30mm depth. a, Sample volume gate axial length 1.5 mm, b, sample volume gate axial length 3mm
J. Biomed.
Eng. 1991, Vol. 13, May
251
Colour Doppler ultrasound accuracy in small vessels: C.R. Deane et al. Table 3 Variations in the PW pulse of the Acuson I.538 with depth (s.v. = 1.5 mm) Depth (mm) 5 30 50 70
assumed lele@yei;
f
Amplitude
Tp
(MHz)
(mv)
(N)
L XL (mm)
3.5 3.5 3.5 3.5
25.5 72.0 148.0 113.0
2.24 2.24 2.24 2.24
1.7 1.7 1.7 1.7
that an increase in sample volume axial achieved by the use of multiple receiving
Tl! e variations in the PW pulse as a function of depth are recorded for the Acuson L.538 in Table 3. (Similar results are found for the S328 transducer.) Note that the depth of interest in PW Doppler refers to the exact location of the sample volume and not to an area of interest as in colour Doppler mode. The increase in peak-to-peak amplitude of the pulse as the depth of interest increases is probably implemented to compensate for signal attenuation. However, the measured amplitude was greatest at a depth of 50mm. At greater depths a slight decrease in amplitude is observed. This behaviour seems to indicate that the Acuson 128 cannot focus its PW Doppler beam over the same range as it covers in Bmode, which implies that some decrease in lateral resolution with depth may occur in PW Doppler, due to the larger beam size. However, no attempt to assess the importance of this phenomenon has been made, since the prime concern of this paper is CDU measurements.
FLOW-RIG
TESTS
ON CDU ACCURACY
Methods A flow rig was constructed using a latex tube of 1.7mm diameter (Tun Abdul Razak Lab., Brickendonbury, UK) (Figure 5). The size of tube was chosen as representative of a segmental/interlobar artery within the kidne . Such vessels are not usually seen in B-mode images E ut are frequently observed in colour Doppler images. The tube was su ported in a tank of water with reticulated foam (Bu Ppren S20) between tank walls and tube and between probe face and tube. The attenuation and velocity characteristics of the foam have been described previously7. The entrance I
Syrmge
pump
(max. 54 ml min-‘)
Microadiuster
Magnetic
stirrer
Figure 5 Diagram of the experimental measurements
252
J. Biomed. Eng. 1991, Vol. 13, May
set-up for the flow-rig
length for all experiments exceeded 120mm, thus ensuring laminar flow and a parabolic flow profile. The fluid used was defibrinated blood with a red blood cell content of 5% (volume). The blood was circulated b means of a twin syringe pump of maximum Kow 54 ml mini, corresponding to a mean blood flow velocity of 0.397ms-’ in the tube. In practice, a flow rate of 20mlmin’ was employed to prolong test periods between recharging the syringes. At this rate, mean and peak flow velocities (assuming parabolic flow) were, respectively, 0.147 and 0.294 m s-l. The transducers were mounted, in turn, on a traversing micro-adjusting slide with the X-Y plane parallel to the line of the tube. The transmit power for PW and colour Doppler were adjusted to obtain signals similar to those observed in vim. Colour Doppler observations were made solely by appraising the image. This is, of necessity, a qualitative measurement. Pulsed Doppler signal strength was obtained by measuring the voltage at the audio output. This measurement can be altered by increasing the gain or audio volume of the pulsed Doppler signal. The controls were set to represent in z1iu0signal strengths when the probe was in line with the centre of the tube and signal strengths were then measured as a percentage of this maximum signal, although the relationship between the audio voltage and the received signal strength is not known. Observations were made at traverse distance intervals of 0.5 mm along the Z axis. Measurements were made at varying transducer-tube distances to examine the erformance of the scanner at a range of depths. For t e L538 probe, the depths were 20, 40 and 60mm; for the S238 probe they were 40,80 and 120mm. While the S328 probe field does extend to 240mm, in practice the attenuation of Doppler signals is such that low flow velocities are not seen at these depths.
R
Results The observations for both probes are shown graphically in Figures 6 and 7. In practice, measurements were started at a transducer position at which the colour Doppler showed no flow and at which the PW signal was at background level. As the probe was brought closer to the tube axis, intermittent colour (shaded areas on Figures 6 and 7) was displayed on the screen and the PW signal increased. In all cases but one (S328 depth 40 mm), the colour intensity and PW signal reached a peak when the robe was in line with the tube. The range in whit K consistent and intense colour was observed is shown as a solid bar. The observations for the L538 transducer show that beam width increases steadily with depth. At 60mm depth, consistent colour is observed over a transverse distance of 3.5mm; if intermittent signals are included, this increases to 5 mm. This contrasts with the results at 20 mm (1.5 and 2.5 mm respectively) and is probably indicative of the difficulty in providing a tightly focused beam width at greater depths. The PW Doppler signals show similar spreading with depth. The observations for the S328 transducer show that
Colour Doppler ultrasound accuracy in small vessels: C.R. Deane et al.
0
Tube section
0
Tube section
*loo% .
SV depth 120 mm
t SV depth 60 mm
.
.
.
.
100%
.
.
. .
.
.
t
.
, .m.
SVdepth40mm
.
SV depth 80 mm
. .
.
.
+
. _...lOOO~
A SV depth
I
I -4
I
I -2
I
I
0
20 mm
.
.
.
A
I
III1
+4
.
(mm1
.ateral distance along Z axis from maximum PW Doppler signal Figure 6 Flow-rig measurements on the IS8 used in colour and PW Doppler mode. The solid and shaded bars represent consistent and intermittent colour Doppler images respectively. The graphical symbols represent the relative amplitude of the PW audio signal as a proportion of the maximum signal obtained. The vessel cross-section is drawn to scale with the horizontal axis
there is a well focused beam at a depth of 80mm, with consistent and intermittent colour over a range of 1.5-2.,5 mm. At 120mm this has only increased slightly. However, at 40 mm, which is in the near field for this transducer, PW and colour Doppler signals are obtained over a wide range (3.5mm - consistent colour, 5mm - intermittent). A fall in PW strength is observed when the probe is in line with the tube. Negative PW Dopplersignals were seen at this point, which could not be excluded despite changing transmit power and gain. The reason for this is not known. The difference in focusing capabilities may be due, in part, to the layout of the transducer elements, the L538 elements being grouped in an area approximately 38 X 1Omm and the S328 elements in an area approximately 28 X 14 mm. DISCUSSION The overall differences between the colour Doppler and PW Doppler modes were not unexpected. The two modes present frequency shift information in different ways; consequently different processing algorithms are applied to the received signals for each modes,‘. The advantages of using longer pulses in
I
. I
-4
SV depth 40 mm * .
*
. +2
.
.
.
.
. I
I
-2
I
0
I
I
+2
I
I
+4
. I
(mm) Lateral distance along Z axis from maximum PW Doppler signal Figure 7 Flow-rig measurements on the S328 used in colour and PW Doppler mode. The symbols are the same as for Figure 6
PW Doppler are two-fold: the effect of the transit time broadening artefact is significantly reduced”; a better SNR is achieved since a larger sample volume means that more signal power is obtained. The increase in pulse duration with depth (by up to 50%) observed when using the Acuson 128 in colour Doppler mode is more surprising, especially as no mention of this phenomenon can be ,found in the manuals accompanying the scanner’.‘. While this secures a better SNR, it does so at the expense of axial resolution. Although the absolute change in axial length of the sample volume (0.5mm) is not excessive, care must be taken when comparing measurements obtained at different depths. The two transducers examined - the ii MHz linear array and the 3.5 MHz phased array - were found to use approximately the same Doppler frequency of 3.Fi MHz. Consequently, no gain in axial resolution can be obtained by changing from the 5 MHz Acuson L538 to the 3.5MHz Acuson S328 transducer. The use of 3.5MHz Doppler pulses instead of 5MHz gives a better, deeper penetration into the tissue and an improved signal-to-noise ratio, again at the expense of a decrease in resolution. A lower frequency in Doppler mode (both in PW and colour) means that the blood velocity estimates (whether in colour flow map or sonograph form) are less susceptible to aliasing, since the Doppler shifts
J. Biomed. Eng. 1991, Vol. 13, May
253
Colour Doppler ultrasound accuracy in small vessels: C.R. Deane et al.
are smaller
(CJ: equation (1)). Conversely, the reduced Doppler shift frequencies will limit the ability to measure low velocities. To measure velocities accurately within a vessel, Dopper angle correction must be possible. The results from the flow-rig test show that colour Doppler imaging can provide an estimate of vessel course, thereby permitting angle correction, but that the error due to beam width is a variable one. For the 5 MHz linear-array transducer, the beam width is well contained at shallow depths (Figure 6), such that the error due to Z-plane thickness is less than 1 mm. If a velocity is measured in the middle of a vessel image of 6mm in length, the error in velocity measurement due to misalignment in the Z plane is <5%. At depths of > 60 mm, the Z-plane thickness can increase to > 3 mm; over the same length of vessel the potential error may be as great as 20%. The possible error due to Z-plane thickness of the phased array sector transducer is greatest at shallow depths but many be reduced to <5% at depths >60 mm. These results show that the transducers are optimized for different depths and that care should be taken when obtaining measurements outside these optimum depth ranges. Similar findings have been obtained in the transmitted beam profile of a 10MHz duplex scanner from another manufacturer’0 and highlight the need for quality assurance of commercial ultrasound systems. CONCLUSION The clinical applications of colour Doppler ultrasound are diverse’ ’ and reported use of the technique is increasing as more systems become available. Whilst the major advantages are qualitative ones, the specific characteristics of colour Doppler are being used to make quantitative measurements. Systems are being sold with little technical information as to how images are achieved and what the practical limitations are. Clearly, even the most basic information, such as Doppler and high-pass filter frequencies, have major ramifications in, for instance, the lowest flow velocities that can be detected. The measurements undertaken in this study are not comprehensive, but they do address some of the technical limitations of one particular colour Doppler ultrasound scanner. With two tests we have shown that colour Doppler sample volume axial length increases with depth and that focusing constraints impose limitations on the accuracy of the colour
254
J. Biomed. Eng. 1991, Vol. 13, May
images of vessels. This information has direct implications for the ability of the system to measure flow velocities, yet it is not readily obtained. Even the Doppler frequencies used on these probes are not supplied to the user by the manufacturer. Purchasers and users of these sophisticated, and systems must be made aware of the expensive, limitations of the machines. Manufacturers should be encouraged to provide sufficient information on the performance, accuracy and safety of their products. ACKNOWLEDGEMENTS We gratefully acknowledge the financial support of the Medical Research Council (N.T.) and the Science and Engineering Research Council (F.F.) and thank Dr S.L.E. Douglas of Ring’s College for help with the flow-rig measurements. REFERENCES 1. Acuson 128 Computed Sonography System: User manual, 2nd edn. California: Acuson Corp. 1988. 2. Acuson 728 Computed Sonography System: Color Doj$br Imaging Option User Manual 3rd edn. California: Acuson Corp. 1989. 3. Peronneau PA, Bournat J-P, Bugnon A, Barbet A, Xhaard M. Theoretical and practical aspects of pulsed Doppler flowmetry: real-time application to the measurement of instantaneous velocity proviles in vitro and in vivo. In: Reneman RS, ed. Cardiovascular Applications of Ultrasound. Amsterdam: North Holland, 1974, 66-84. 4. Kristoffersen K. Optimal receiver filtering in pulsed Doppler ultrasound blood velocity measurements. IEEE Trans Ultra-son Ferroelec Freq Contr 1986; 31: 51-8. 5. Wells PNT. Biomedical Ultrasonics. London: Academic Press, 1977. 6. Halberg LI, Thiele KE. Extraction of blood flow information using Dopper-shifted ultrasound. Hewlett-Packard J 1986; 37: 35-40. 7. Lerski RA, Duggan TC, Christie J. A simple tissue-like ultrasound phantom material. BrJRadioll982; 55: 156-7. 8. Burns PN. The physical principles of Doppler ultrasound and spectral analysis. J Clin Ultrasound 1987; 15: 567-90. 9. Kasai C, Namekawa N, Koyano A, Omoto R. Real-time blood flow imaging using an autocorrelation technique. IEEE Trans Sonics Ultrason 1985; 32: 458-64. 10. Oates CP, Williams Ed, McHugh MI. The use of a Diasonics DRP400 duplex ultrasound scanner to measure volume flow in arterio-venous Bstulae in patients undergoing haemodialysis: an analysis of measurement uncertainties. Ultrasound Med Biol 1990; 16: 571-9. 11. Merritt CR. Doppler color flow imaging. J Clin Ultrasound 1987; 15: 591-7.
velocity
N. Thomas and V.C. Roberts
Department of Medical Physics and Medical Engineering, King’s College School of Medicine and Dentistry, Dulwich Hospital, East Dulwich Grove, London SE22 8PT, UK
ABSTRACT ColourDoppler ultrasound offers thepossibility of imaging small vessels not visible by B-mode alone. The colour Doppler image of velocities allows the course of small vessels to be imaged in the X-Yplane of the scan provided the Doppler s E 1so suI?’ nent magnitude. This permits alignments of the Dapple r cursor, allowing angle correction to f;qe uenqhlfi’f provide true velocity measurements from the Doppler shiJi obtained. Before attempting to make velocity measurements, however, it is essential to be aware of the possible error in the Z plane caused by the thickness of the Doppler sample volume. To quantzfi this source of error, hydrophone and flow-rig measurements were peformed on an Acuson 128 colour Doppler scanner with both 5MHz linear-array and 3.5MHz phased-array transducers. Measurements of the transmittedpulses using a point hydrophone showed that both probes employ approximately 3.5MHz Dopplerpulses (in both colour andpulsed Doppler modes). The two transducers have the same axial resolution. In colour Doppler mode the axial length of the sample volume increases automatically with depth by up to 0.5mm. Measurements of colour and pulsed Doppler signal strength were obtained in a controlledJl ow rig. Both transducers produced accurate colourJow images of the phantom at their optimum depths; flow velocity errors due to Z-plane thickness are < 5%. There was, however, substantial error outside these optimum conditions (up to 20%). Keywords: Colour Doppler
ultrasound,
hydrophone
measurements,
INTRODUCTION Colour Dop ler ultrasound (CDU) scanners have been availab Pe for vascular investigations for over 2 years. The use of colour Doppler images to scan an area for Doppler frequency shifts offers advantages when compared with conventional duplex Doppler scanning, where only a specific volume or ‘gate’ is analysed at any time: it provides rapid assessment as to whether a structure under investigation is vascularized or not, highlights areas of high or abnormal flow; and can assist pulsed wave Doppler (PW) investigations by allowing rapid positioning of the PW sample volume. CDU also offers potential quantitative information in showing the course of small vessels not visible in Bmode images alone. This permits PW angle correction to allow measurement of arterial and venous velocities by PW using the classic Doppler equation: fi,=
2v cost3 ---fc
(1)
cb
where ft, is the Doppler frequency shift measured, u is the velocity of the blood, 8 is the angle to be corrected between the ultrasound beam and the vessel, ct is the velocity of sound in tissue and fcis the transmitted ultrasound frequency. The accuracy of the angular correction is, however, largely dependent Correspondence
to: CR.
tests, Acuson
128
on the resolution of the colour Doppler image. This is ap arent in the plane on the screen (the X-Y plane 7 . What is not readily appreciated is the ambiguity in the Z plane caused by the beam and sampling width of the scanner when in colour Doppler mode, as illustrated in Figure 1. In this paper we present work undertaken on an Acuson 128 colour Doppler scanner. The aim was to investigate the way in which the colour Doppler scanning is accomplished and to attem t to quantify the errors obtained when using a co Pour Doppler image to provide the pulsed Doppler beam-vessel angle correction, 0. The work falls into two parts:
\ t!L
1. Using a hydrophone,
examination
of the emitted
Beam width
-I
iy
Vessel
\ \J/’ Colour Doppler
sample volume
Figure 1 2-D Doppler
Deane
0 1991 Butterworth-Heinemann 0141-:i42.5/91/030249-06
flow-rig
image representing
a 3-D blood vessel
for BES J. Biomed.
Eng. 1991, Vol. 13, May
249
Colour Doppler ultrasound accuracy in small vessek CR. Deane et al.
signals in B-mode, pulsed Doppler and colour Doppler modes. 2. Assessment of the accuracy of colour and pulsed Doppler measurements made on a controlled flow through a latex tube of known dimensions.
Gould 4050
w
oscilloscope
Absorbing foam
The Acuson 128 (Acuson, Mountain View, CA, USA) is a commercial ultrasound scanner with both pulsed Doppler and colour Doppler facilities. The transducers used during these studies were: Acuson I538 linear-array transducer, nominally 5 MHz with an aperture of 38 mm and a maximum depth of image of 80mm. S328 phased-array ‘sector’ transducer, 0 Acuson nominally 3.5 MHz with a maximum depth of image of 240mm.
IEEE 488
0
Very little technical information is available to users of the scanner; the information provided is primarily intended for operating the machine”L. The CDU investigation area may be altered; in the case of the I538 probe, from full frame to an area of 10 X 10mm. When the CDU area is reduced the frame rate is increased slightly and the colour pixel size becomes smaller. Similar changes occur with the S328 probe.
HYDROPHONE MEASUREMENT SOFTHE TRANSMITTED PULSE The resolution of the colour Doppler image is constrained by the size of the sample volume. For pulsed Doppler (including colour Doppler) the axial resolution of the sample volume, L,, is dictated by the duration of the pulse (number of periods within the pulse) and the filter characteristics of the receiver”T4 according to: =
L
experimental
set-up
for
the
pulses for the colour Doppler signal linear-array transducer are shown in Figure 3a-c for three different depths (0, 30 and 70 mm respectively). The Doppler frequency emitted is 3.5 MHz in spite of the B-mode, as specified for the transducer, operating at 5MHz. The measured values of both amplitude and pulse length are given in Table 7. The axial length of the sample volume due solely to the duration of the emitted pulse, L,,, was calculated according to equation (2). In practice, the axial length is also dependent on T,; however, absolute changes to Lxt contribute direct1 to L,. The observed increase in (peak-to-peak r amplitude from 110 to 200 mV is implemented to counteract the effect of attenuation on the signal-to-noise ratio (SNR). H owever, the increase in the number of periods emitted (from 4 to 7) was quite unexpected. The equivalent change in L,, is 50% (c$ Table 7). The result is that the axial resolution is reduced with depth. The same set of measurements have been performed using the 3.5 MHz phased-array transducer (S328). In this case the B- and colour Doppler-modes were found to employ the same centre frequency of 3.5 MHz. However, a similar reduction in colour Doppler axial resolution for an increase in depth of on the 5MHz
where Tp is the duration of the emitted pulse, T, is the duration of the receiver filter impulse response and cb is the average veloci of sound in blood (approximately 1570m s-‘) (re 9erence 5). It was not possible to investigate directly the characteristics of the receiver filter. However, to assess the sources of error arising from the contribution of pulse duration to the axial length of the sample volume, we measured T,, for both transducers in all modes at a range of depths. Methods
The transmitted pulses were measured in a water bath containing distilled water. The transducer in question and a 1 mm PZT point hydrohone (Mao-8X, A0 Williams, Malmesbu UK) were mounted on a fixture at a variable 7 ittance apart. After a manual alignment to secure maximum output neither the transducer nor the hydrophone was moved. A block diagram of the experimental set-up is given in
250
of the
phone was sampled directly using a Gould 4050 digital storage oscilloscope, which has a resolution of 8 bits and a sampling frequency of 100MHz. The sampled signal was transferred to a Compaq 386 personal computer, via an IEEE-488 interface, for processing and dis lay. from the The two trans cfucers were controlled Acuson 128 colour Doppler scanner. To obtain CDU pulses at a range of depths, the area of colour Doppler display on the B-mode image was reduced to a minimum. On an Acuson 128 the smallest area possible in colour Doppler studies is a square of 1 x 1 cm. All depths given in the following refer to the top of this square area. An absorbent foam was placed in the water bath to prevent reflections from the side of the bath interfering with the original waveform.
The recorded
2
Figure 2. The high frequency
Block diagram measurements
Results
dTp+ 7;)
x
Figure 2 hydrophone
signal from the point hydro-
J. Biomed. Eng. 1991, Vol. 13, May
CohT Doppler ultrasound accuracy in small vessels: C.R. Deane et al. Table 1
0.151
O.lj
7
0.051
d f g
Measurements on the Acuson L53H in colour Doppler mode
Depth
f
Amplitude
(mm) 0 30 70
(MHz) 3.5 3.5 3.5
(mv)
(?s,
[Irn)
100
I .32 I .70
1.0 1.3 I .5
200 190
I.95
01 Table 2
5
-0.051
Measurements
on the Acuson S328 in colour Doppler mode
Depth
f
Amplitude
(mm)
(MHz)
(mV)
2s)
I (zn)
6: 180
3.2 3.2
200 8.5 200
1.84 I .43 1.93
I.1 1.4 1.5
-O.l!
-0.15’ 0
1
2
3
4.5
i
a 1.5 to 30 mm, the number of emitted periods changes from 8 to 15, as illustrated in Figure 4u and b. These measurements, taken on the Acuson L538 transducer, are equivalent to a change in Lxt from 1.7 to 3.4mm, and demonstrate the direct relationship between pulse duration and sample volume axial length. A similar behaviour was observed for the 3.5 MHz At sample gate settings phased-array transducer. greater than this (up to 30mm), there is no further increase in the number of transmitted periods. It is
Time (x 10m6 s)
a 0.15’ 0.1 p
0.05t
0” 3 C
0’
g -0.05-
-0.q
-0.15L 0
7
8
Time (~10~~ s)
b
a 0.151
-0.02 t -0.03/
0.11
-0.04b 2
0.05
$ 2 “a E
Time (~10~~ s)
a
o;_
0.04
4-o.05t I
0.03-
-0.1; -0.15 i 0
C
f
s
Acuson
L.538
Time (x 10v6 s)
Figure 3 Emitted colour Doppler pulse of the transducer focused at: a, Omm; b, 30mm; c, 70 mm
the region of imaging was measured as shown in Table 2. Since the two transducers reduce almost identical Doppler pulses (compare $ ables 7 and 2) the S328 pulse is not shown. the transducers’ corresponding For comparison, PW pulses were studied as well. In PW mode, the sample gate is under direct user control; the axial length may be varied from 1.5 to 30 mm. In the range
-0.031 -0.04&
b
Lo
1
2
._
1~
~~~
3
4
5
6
7
8
Time (~10~~ s)
Figure 4 PW pulse of the L538 recorded at 30mm depth. a, Sample volume gate axial length 1.5 mm, b, sample volume gate axial length 3mm
J. Biomed.
Eng. 1991, Vol. 13, May
251
Colour Doppler ultrasound accuracy in small vessels: C.R. Deane et al. Table 3 Variations in the PW pulse of the Acuson I.538 with depth (s.v. = 1.5 mm) Depth (mm) 5 30 50 70
assumed lele@yei;
f
Amplitude
Tp
(MHz)
(mv)
(N)
L XL (mm)
3.5 3.5 3.5 3.5
25.5 72.0 148.0 113.0
2.24 2.24 2.24 2.24
1.7 1.7 1.7 1.7
that an increase in sample volume axial achieved by the use of multiple receiving
Tl! e variations in the PW pulse as a function of depth are recorded for the Acuson L.538 in Table 3. (Similar results are found for the S328 transducer.) Note that the depth of interest in PW Doppler refers to the exact location of the sample volume and not to an area of interest as in colour Doppler mode. The increase in peak-to-peak amplitude of the pulse as the depth of interest increases is probably implemented to compensate for signal attenuation. However, the measured amplitude was greatest at a depth of 50mm. At greater depths a slight decrease in amplitude is observed. This behaviour seems to indicate that the Acuson 128 cannot focus its PW Doppler beam over the same range as it covers in Bmode, which implies that some decrease in lateral resolution with depth may occur in PW Doppler, due to the larger beam size. However, no attempt to assess the importance of this phenomenon has been made, since the prime concern of this paper is CDU measurements.
FLOW-RIG
TESTS
ON CDU ACCURACY
Methods A flow rig was constructed using a latex tube of 1.7mm diameter (Tun Abdul Razak Lab., Brickendonbury, UK) (Figure 5). The size of tube was chosen as representative of a segmental/interlobar artery within the kidne . Such vessels are not usually seen in B-mode images E ut are frequently observed in colour Doppler images. The tube was su ported in a tank of water with reticulated foam (Bu Ppren S20) between tank walls and tube and between probe face and tube. The attenuation and velocity characteristics of the foam have been described previously7. The entrance I
Syrmge
pump
(max. 54 ml min-‘)
Microadiuster
Magnetic
stirrer
Figure 5 Diagram of the experimental measurements
252
J. Biomed. Eng. 1991, Vol. 13, May
set-up for the flow-rig
length for all experiments exceeded 120mm, thus ensuring laminar flow and a parabolic flow profile. The fluid used was defibrinated blood with a red blood cell content of 5% (volume). The blood was circulated b means of a twin syringe pump of maximum Kow 54 ml mini, corresponding to a mean blood flow velocity of 0.397ms-’ in the tube. In practice, a flow rate of 20mlmin’ was employed to prolong test periods between recharging the syringes. At this rate, mean and peak flow velocities (assuming parabolic flow) were, respectively, 0.147 and 0.294 m s-l. The transducers were mounted, in turn, on a traversing micro-adjusting slide with the X-Y plane parallel to the line of the tube. The transmit power for PW and colour Doppler were adjusted to obtain signals similar to those observed in vim. Colour Doppler observations were made solely by appraising the image. This is, of necessity, a qualitative measurement. Pulsed Doppler signal strength was obtained by measuring the voltage at the audio output. This measurement can be altered by increasing the gain or audio volume of the pulsed Doppler signal. The controls were set to represent in z1iu0signal strengths when the probe was in line with the centre of the tube and signal strengths were then measured as a percentage of this maximum signal, although the relationship between the audio voltage and the received signal strength is not known. Observations were made at traverse distance intervals of 0.5 mm along the Z axis. Measurements were made at varying transducer-tube distances to examine the erformance of the scanner at a range of depths. For t e L538 probe, the depths were 20, 40 and 60mm; for the S238 probe they were 40,80 and 120mm. While the S328 probe field does extend to 240mm, in practice the attenuation of Doppler signals is such that low flow velocities are not seen at these depths.
R
Results The observations for both probes are shown graphically in Figures 6 and 7. In practice, measurements were started at a transducer position at which the colour Doppler showed no flow and at which the PW signal was at background level. As the probe was brought closer to the tube axis, intermittent colour (shaded areas on Figures 6 and 7) was displayed on the screen and the PW signal increased. In all cases but one (S328 depth 40 mm), the colour intensity and PW signal reached a peak when the robe was in line with the tube. The range in whit K consistent and intense colour was observed is shown as a solid bar. The observations for the L538 transducer show that beam width increases steadily with depth. At 60mm depth, consistent colour is observed over a transverse distance of 3.5mm; if intermittent signals are included, this increases to 5 mm. This contrasts with the results at 20 mm (1.5 and 2.5 mm respectively) and is probably indicative of the difficulty in providing a tightly focused beam width at greater depths. The PW Doppler signals show similar spreading with depth. The observations for the S328 transducer show that
Colour Doppler ultrasound accuracy in small vessels: C.R. Deane et al.
0
Tube section
0
Tube section
*loo% .
SV depth 120 mm
t SV depth 60 mm
.
.
.
.
100%
.
.
. .
.
.
t
.
, .m.
SVdepth40mm
.
SV depth 80 mm
. .
.
.
+
. _...lOOO~
A SV depth
I
I -4
I
I -2
I
I
0
20 mm
.
.
.
A
I
III1
+4
.
(mm1
.ateral distance along Z axis from maximum PW Doppler signal Figure 6 Flow-rig measurements on the IS8 used in colour and PW Doppler mode. The solid and shaded bars represent consistent and intermittent colour Doppler images respectively. The graphical symbols represent the relative amplitude of the PW audio signal as a proportion of the maximum signal obtained. The vessel cross-section is drawn to scale with the horizontal axis
there is a well focused beam at a depth of 80mm, with consistent and intermittent colour over a range of 1.5-2.,5 mm. At 120mm this has only increased slightly. However, at 40 mm, which is in the near field for this transducer, PW and colour Doppler signals are obtained over a wide range (3.5mm - consistent colour, 5mm - intermittent). A fall in PW strength is observed when the probe is in line with the tube. Negative PW Dopplersignals were seen at this point, which could not be excluded despite changing transmit power and gain. The reason for this is not known. The difference in focusing capabilities may be due, in part, to the layout of the transducer elements, the L538 elements being grouped in an area approximately 38 X 1Omm and the S328 elements in an area approximately 28 X 14 mm. DISCUSSION The overall differences between the colour Doppler and PW Doppler modes were not unexpected. The two modes present frequency shift information in different ways; consequently different processing algorithms are applied to the received signals for each modes,‘. The advantages of using longer pulses in
I
. I
-4
SV depth 40 mm * .
*
. +2
.
.
.
.
. I
I
-2
I
0
I
I
+2
I
I
+4
. I
(mm) Lateral distance along Z axis from maximum PW Doppler signal Figure 7 Flow-rig measurements on the S328 used in colour and PW Doppler mode. The symbols are the same as for Figure 6
PW Doppler are two-fold: the effect of the transit time broadening artefact is significantly reduced”; a better SNR is achieved since a larger sample volume means that more signal power is obtained. The increase in pulse duration with depth (by up to 50%) observed when using the Acuson 128 in colour Doppler mode is more surprising, especially as no mention of this phenomenon can be ,found in the manuals accompanying the scanner’.‘. While this secures a better SNR, it does so at the expense of axial resolution. Although the absolute change in axial length of the sample volume (0.5mm) is not excessive, care must be taken when comparing measurements obtained at different depths. The two transducers examined - the ii MHz linear array and the 3.5 MHz phased array - were found to use approximately the same Doppler frequency of 3.Fi MHz. Consequently, no gain in axial resolution can be obtained by changing from the 5 MHz Acuson L538 to the 3.5MHz Acuson S328 transducer. The use of 3.5MHz Doppler pulses instead of 5MHz gives a better, deeper penetration into the tissue and an improved signal-to-noise ratio, again at the expense of a decrease in resolution. A lower frequency in Doppler mode (both in PW and colour) means that the blood velocity estimates (whether in colour flow map or sonograph form) are less susceptible to aliasing, since the Doppler shifts
J. Biomed. Eng. 1991, Vol. 13, May
253
Colour Doppler ultrasound accuracy in small vessels: C.R. Deane et al.
are smaller
(CJ: equation (1)). Conversely, the reduced Doppler shift frequencies will limit the ability to measure low velocities. To measure velocities accurately within a vessel, Dopper angle correction must be possible. The results from the flow-rig test show that colour Doppler imaging can provide an estimate of vessel course, thereby permitting angle correction, but that the error due to beam width is a variable one. For the 5 MHz linear-array transducer, the beam width is well contained at shallow depths (Figure 6), such that the error due to Z-plane thickness is less than 1 mm. If a velocity is measured in the middle of a vessel image of 6mm in length, the error in velocity measurement due to misalignment in the Z plane is <5%. At depths of > 60 mm, the Z-plane thickness can increase to > 3 mm; over the same length of vessel the potential error may be as great as 20%. The possible error due to Z-plane thickness of the phased array sector transducer is greatest at shallow depths but many be reduced to <5% at depths >60 mm. These results show that the transducers are optimized for different depths and that care should be taken when obtaining measurements outside these optimum depth ranges. Similar findings have been obtained in the transmitted beam profile of a 10MHz duplex scanner from another manufacturer’0 and highlight the need for quality assurance of commercial ultrasound systems. CONCLUSION The clinical applications of colour Doppler ultrasound are diverse’ ’ and reported use of the technique is increasing as more systems become available. Whilst the major advantages are qualitative ones, the specific characteristics of colour Doppler are being used to make quantitative measurements. Systems are being sold with little technical information as to how images are achieved and what the practical limitations are. Clearly, even the most basic information, such as Doppler and high-pass filter frequencies, have major ramifications in, for instance, the lowest flow velocities that can be detected. The measurements undertaken in this study are not comprehensive, but they do address some of the technical limitations of one particular colour Doppler ultrasound scanner. With two tests we have shown that colour Doppler sample volume axial length increases with depth and that focusing constraints impose limitations on the accuracy of the colour
254
J. Biomed. Eng. 1991, Vol. 13, May
images of vessels. This information has direct implications for the ability of the system to measure flow velocities, yet it is not readily obtained. Even the Doppler frequencies used on these probes are not supplied to the user by the manufacturer. Purchasers and users of these sophisticated, and systems must be made aware of the expensive, limitations of the machines. Manufacturers should be encouraged to provide sufficient information on the performance, accuracy and safety of their products. ACKNOWLEDGEMENTS We gratefully acknowledge the financial support of the Medical Research Council (N.T.) and the Science and Engineering Research Council (F.F.) and thank Dr S.L.E. Douglas of Ring’s College for help with the flow-rig measurements. REFERENCES 1. Acuson 128 Computed Sonography System: User manual, 2nd edn. California: Acuson Corp. 1988. 2. Acuson 728 Computed Sonography System: Color Doj$br Imaging Option User Manual 3rd edn. California: Acuson Corp. 1989. 3. Peronneau PA, Bournat J-P, Bugnon A, Barbet A, Xhaard M. Theoretical and practical aspects of pulsed Doppler flowmetry: real-time application to the measurement of instantaneous velocity proviles in vitro and in vivo. In: Reneman RS, ed. Cardiovascular Applications of Ultrasound. Amsterdam: North Holland, 1974, 66-84. 4. Kristoffersen K. Optimal receiver filtering in pulsed Doppler ultrasound blood velocity measurements. IEEE Trans Ultra-son Ferroelec Freq Contr 1986; 31: 51-8. 5. Wells PNT. Biomedical Ultrasonics. London: Academic Press, 1977. 6. Halberg LI, Thiele KE. Extraction of blood flow information using Dopper-shifted ultrasound. Hewlett-Packard J 1986; 37: 35-40. 7. Lerski RA, Duggan TC, Christie J. A simple tissue-like ultrasound phantom material. BrJRadioll982; 55: 156-7. 8. Burns PN. The physical principles of Doppler ultrasound and spectral analysis. J Clin Ultrasound 1987; 15: 567-90. 9. Kasai C, Namekawa N, Koyano A, Omoto R. Real-time blood flow imaging using an autocorrelation technique. IEEE Trans Sonics Ultrason 1985; 32: 458-64. 10. Oates CP, Williams Ed, McHugh MI. The use of a Diasonics DRP400 duplex ultrasound scanner to measure volume flow in arterio-venous Bstulae in patients undergoing haemodialysis: an analysis of measurement uncertainties. Ultrasound Med Biol 1990; 16: 571-9. 11. Merritt CR. Doppler color flow imaging. J Clin Ultrasound 1987; 15: 591-7.