Professional Documents
Culture Documents
1221–1227, 1999
Copyright © 1999 World Federation for Ultrasound in Medicine & Biology
Printed in the USA. All rights reserved
0301-5629/99/$–see front matter
PII S0301-5629(99)00083-6
● Original Contribution
Abstract—In this paper, we present an automatic beam-vector (Doppler) angle and flow velocity measurement
method and implement it in pulsatile flow measurements using a clinical Doppler ultrasound system. In current
clinical Doppler ultrasound flow velocity measurements, the axis of the blood vessel needs to be set manually on
the B-scan image to enable the estimation of the beam-vector angle and the beam-vector angle corrected flow
velocity (the actual flow velocity). In this study, an annular array transducer was used to generate a conical-
shaped and symmetrically focused ultrasound beam to measure the flow velocity vectors parallel and perpen-
dicular to the ultrasound beam axis. The beam-vector angle and flow velocity is calculated from the mode
frequency (fd) and the maximum Doppler frequency (fmax) of the Doppler spectrum. We develop a spectrum
normalization algorithm to enable the Doppler spectrum averaging using the spectra obtained within a single
cardiac cycle. The Doppler spectrum averaging process reduces the noise level in the Doppler spectrum and also
enables the calculation of the beam-vector angle and flow velocity for pulsatile flows to be measured. We have
verified the measurement method in vivo over a wide range of angles, from 52° to 80°, and the standard deviations
of the measured beam-vector angles and flow velocities in the carotid artery are lower than 2.2° and 12 cm/s
(about 13.3%), respectively. © 1999 World Federation for Ultrasound in Medicine & Biology.
Key Words: Flow velocity measurement, Doppler spectrum, Beam-vector angle, Spectrum averaging,
Annular array
1221
1222 Ultrasound in Medicine and Biology Volume 25, Number 8, 1999
PRINCIPLE
Fig. 1. The schematic diagram of the beam-vector angle and
flow velocity measurement using an annular array transducer. In classic Doppler, the relationship between the
The transverse, longitudinal components of the flow velocity, flow velocity v and the mode frequency fd of the Doppler
and the beam-vector angle can be measured. VA and VB are two spectrum is:
arbitrary flow vectors with the same beam-vector angle.
2䡠v
fd ⫽ 䡠 cos (1)
ence of noise and signal variations in the Doppler spec-
trum, the Doppler spectrum needs to be averaged many
times to reduce the spectral noise. Figure 2 shows the where is the wavelength of acoustic wave, and is the
variation in the calculated beam-vector angle vs. the measured beam-vector angle. Sometimes, fd is also re-
number of spectral lines in the averaging. The standard ferred to as the mean frequency of the Doppler spectrum
deviations of the calculated angles are less than 2°, when (Tortoli et al. 1995), rather than the mode frequency used
more than 15 spectra are used in the averaging. Because in this article. Theoretically, when the center of the
the flow velocity of a constant flow is time-invariant, the sample volume is much smaller than the diameter of a
Doppler spectra can be directly averaged, whereas in conduit and is positioned at the center of the conduit, the
pulsatile flow conditions, the flow velocities are time- mean frequency equals to the mode frequency (Law et al.
1991).
For a converging ultrasound beam crossed by a flow
line, a multiplicity of ray-to-flow line angles can be said
to exist. Instead of a single frequency shift, a broad-range
frequency shift around fd would be observed in a mea-
sured Doppler spectrum. This broadening effect has been
described as transverse Doppler theory by Newhouse et
al. (1987) and is described by the following equation:
2 W
Bd ⫽ 䡠 䡠 sin (2)
F
冑冋 册 冋
(7)
册
2
䡠F 2 f max⫺2 2
⫽ 䡠f ⫹ 䡠 共 f max ⫺ f d兲 (4)
2 d W
The scaling factor R can be also determined from the
ratio of the energy peak Doppler frequency fd⫺1 兾fd⫺2.
⫽ tan⫺1 冋 f max ⫺ f d 2 䡠 F
fd
䡠
W 册 (5) Because the maximum Doppler shift frequency fmax
would be less noisy, the ratio of fmax⫺1兾fmax⫺2 is used to
determine the scaling factor R. Thus, a set of Doppler
Equations (4) and (5) have been theoretically derived and spectra can be normalized to a specific Doppler spectrum
experimentally verified in constant flow measurements by a scaling factor R before the spectrum-averaging
(Lee and Chiang 1999). process.
The main problem associated with using eqns (4)
and (5) is that the Doppler spectrum needs averaging to METHODS
reduce the noise level, and then to determine the stable
values of fd and fmax. Averaging a consecutive set of Flow phantom measurement
Doppler spectra is straightforward in constant flow mea- Doppler systems. A color Doppler ultrasound sys-
surements, but requires ECG gating or some other means tem, GE-Vingmed CFM-750 (GE-VingMed Sound,
for pulsatile flow measurements. Figure 2 shows the Horten, Norway), with a four-element annular array
variation of the calculated beam-vector angle vs. the transducer, operating at 2.5 MHz in Doppler mode and
number of lines of spectral data averaged. The beam 3.25 MHz in B-scan, 1.47 cm in diameter and 7.8 cm in
vessel angle was set at 70° and the standard deviation focal length, was used in the flow phantom experiments.
was less than 2° after 15 spectra lines were averaged. The The length of sample volume was kept at 1.5 mm long
Doppler spectrum-averaging process reduces the noise and was set at the focus. The pulsed repetition frequency
level and fluctuations in the spectrum, and improves the was 10 kHz. Each Doppler spectrum calculation time
accuracy of the measurement. However, in pulsatile flow (including sampling and computation time) in the CFM-
measurements, the flow velocities vary with time; thus, 750 is less than 8 ms.
the spectrum cannot be averaged directly as in constant
flow measurements. Flow phantom. We used a UHDC (Shelley Medical
To reduce the noise level and fluctuations in the Imaging Technologies, London, Ontario, Canada) flow
Doppler spectrum obtained from a pulsatile flow mea- phantom to generate pulsatile flows. The flow velocity at
surements, a Doppler spectrum normalization and aver- the systolic peak of pulsatile flow was set at 33 cm/s in
aging algorithm is developed. The normalization of the a latex conduit with a 9.5-mm internal diameter. The
Doppler spectrum is based on the assumption that the conduit was 1 m long and was immersed horizontally in
beam-vector angle is unchanged during the fast systolic a water tank. The UHDC blood mimic fluid was used as
region of the cardiac cycle (about 0.12 s). Therefore, for the scattering source in the flow. The density of the fluid
pulsatile flows measured at a fixed beam-vector angle, was 1.01 g/mL, and the dynamic viscosity was 1.34 cp (1
the maximum Doppler shift frequency fmax can be sim- cp ⫽ 10⫺2 g/s 䡠 cm ).
plified from eqn (3): Signal processing. The maximum Doppler fre-
quency fmax is defined as the intersection of the Doppler
f max ⫽ 䡠 ⌬ (6) spectrum, and an empirically determined threshold is set
at ⫺6 dB. The setting of the threshold for fmax should be
where the constant empirically determined for each Doppler ultrasound sys-
tem, because each system has its own Doppler gain
⌬⫽
2
冉cos ⫹
W
2䡠F
䡠 sin 冊 curve. In this experiment, the threshold was determined
by calibrating a set of estimated beam-vector angles to a
1224 Ultrasound in Medicine and Biology Volume 25, Number 8, 1999
set of known angles (e.g., 50°, 60°, 70°) set in UHDC and Fig. 3D represents the beam-vector angle corrected
flow phantom. Different thresholds (including ⫺6 dB, flow velocities calculated by mode frequencies at the
⫺8 dB, ⫺10 dB, ⫺15 dB and ⫺20 dB) have been systolic peak. In Fig. 3B–D, the solid lines represent the
empirically tested by a controlled flow phantom system. mean values, and the error bars represent the standard
In measuring the flow velocity of the pulsatile flow, deviations in 12 trials. The standard deviations of calcu-
a consecutive set of the Doppler spectra was obtained in lated flow velocities in Fig. 3D are lower than 2.8 cm/s,
the systolic cycle (as shown in the white dotted rectan- and the standard deviations of calculated beam-vector
gular window in Fig. 3A). These measured spectra were angles in Fig. 3C are lower than 2.2°. In Fig. 3B, fmax and
normalized to the Doppler spectrum at the systolic peak fd at 85° are overestimated, because the measured Dopp-
and were averaged to obtained a less noisy Doppler ler spectrum is affected significantly by the Doppler
spectrum. The reasons why we chose the Doppler spectra system’s wall filter and most of the Doppler spectra are
in the systolic cycle were that (1) the spectra of fast flows filtered out. Therefore, fd are overestimated and (fmax ⫺
would be less affected by the wall filter of the Doppler fd) are underestimated, so that the beam-vector angle will
system, and (2) to insure the measured spectra to be in be underestimated, as shown in Fig. 3C.
the laminar flow.
Figure 3A shows the Doppler spectrum of a pulsa- In vivo measurements
tile flow generated by a UHDC flow phantom. The The measurements were conducted at different
systolic peak flow velocity was set at 33 cm/s. The beam-vessel angles between 52° and 80° in 4° incre-
beam-vector angle was set at 50°. A consecutive set of ments. Figure 4B and C show the calculated beam-vector
Doppler spectra obtained during the systolic cycle was angles and the angle corrected flow velocities of a 25-y-
normalized and averaged to estimate the beam-vector old male in six trials. In Fig. 4B and C, the solid lines
angle and the flow velocity at the systolic peak. represent the mean values, and error bars represent the
standard deviations. The standard deviations of the
In vivo measurements beam-vector angles are lower than 2.2°, as shown in Fig.
For in vivo measurements, a different type of annu- 4B, and standard deviations of flow velocities are lower
lar array transducer, four-element, operating at 3.5 MHz than 12 cm/s, as shown in Fig. 4C.
in Doppler mode and 4 MHz in B-scan mode, 1.15 cm in Figure 5A–C shows the traces of the flow velocities
diameter, and 4.7 cm in focal length, was used to mea- of a human subject’s carotid artery measured, respec-
sure the carotid artery of a 25-y-old male subject, from tively, at 56°, 68° and 76° beam-vessel angle settings.
52° to 80° beam-vessel angle. The beam-vector angle The calculated beam-vector angles are 57.5°, 69° and
and flow velocity were calculated in the chosen part in 72.8°, respectively. The three traces represent the flow
the measured spectra (as shown in Fig. 4A). The GE- velocities calculated without angle correction, with man-
VingMed 750 can provide 18 spectra within 0.12 s in the ually set beam-vessel angle correction, and with the
chosen part of the measured spectra. automatic beam-vector angle correction (our method).
In a human subject’s carotid artery measurements, The major limitation in spectral Doppler measure-
the longitudinal direction of the carotid artery was first ments is caused by the Doppler system’s wall filter. It
manually set using a vessel axis marker on the B-scan causes difficulties in measuring slow flow or at large
ultrasound image. The vessel axis and the boundary of beam-vector angles. The existence of the wall filter is
the blood vessel are highlighted by solid line in Fig. 4A. essential, but it inevitably filters the lower frequency
The Doppler sample volume was set at the center of the portion of the Doppler spectrum. We can observe, in Fig.
blood vessel and kept at the minimum 1.5 mm. The value 5C, that flows measured at large beam-vector angles or
of the reference beam-vessel angle is determined visually measured during their slow flow region would be af-
by physicians by aligning the vessel axis marker along fected by the wall filter. Therefore, the spectral Doppler
the blood vessel axis. This process may incur some method is not suitable for large beam-vector angles (larg-
human errors in determining the value of the reference er than 80° in this case) and slow flow velocity (lower
beam-vessel angle. than 18 cm/s) measurements.
RESULTS DISCUSSION
Flow phantom measurements The best measurement position is at the focus where
Figure 3B represents the fmax and fd determined the beam has the narrowest width and highest intensity;
from the spectrum normalization and averaging process the effect of the lateral transit time broadening will be
from 45° to 85° beam-vessel angles (reference angles). most salient, and the signal will also be most sensitive. In
Figure 3C represents the calculated beam-vector angles, modern ultrasound systems, the multiple zone focusing is
Spectral width Doppler in pulsatile flow ● B.-R. LEE et al. 1225
CONCLUSIONS
The major assumption of the presented Doppler
spectrum normalization algorithm is that the beam-vec-
tor angle remains unchanged during the chosen part of
the cardiac cycle. Therefore, a consecutive set of Dopp-
ler spectra can be normalized and averaged, even though
these Doppler spectra are obtained from different flow
velocities. The time in the chosen part of the cardiac
cycle is less than 0.12 s. Thus, it could be easily imple-
mented in real time operation with appropriate imple-
mentation.
It is worth noting that this simple method only Hein IA. 3-D flow velocity vector estimation with a triple-beam lens
transducer experimental results. IEEE Trans Ultra Ferro Freq Con
measures the sign and magnitude of the beam-vector 1997;44:85– 89.
angle ; that is, the magnitude of the transverse flow Law YF, Bascom PAJ, Johnston KW, Vaitkus P, Cobbold RSC.
velocity vector ⱍv⬜ⱍ , and the longitudinal flow velocity Experimental study of the effects of pulsed Doppler sample volume
size and position on the Doppler spectrum. Ultrasonics 1991;29:
vector v//, as shown in Fig. 1. Thus, two arbitrary flows 404 – 410.
VA and VB with the same magnitude of beam-vector Lee BR, Chiang HK, Kuo CD, Lin WL, Lee SK. Doppler angle and
angle A and B cannot be distinguished by using this flow velocity estimations using the classic and transverse Doppler
effects. IEEE Trans Ultra Ferro Freq Con 1999;46:252–256.
method. However, this limitation does not affect mea- Newhouse VL, Censor D, Vontz T, Cisneros JA, Goldberg BB. Ultra-
suring the flow velocity in regular blood vessels, in sound Doppler probing of flows transverse with respect to beam
which the blood flows in only one direction. Therefore, axis. IEEE Trans Biomed Eng 1987;34:779 –789.
Newhouse VL, Dickerson K, Cathignol SD, Chapelon JY. Three-
the application of this simple method could successfully dimensional vector flow estimation using two transducers and spec-
measure the beam-vector angle and flow velocity in tral width. IEEE Trans Ultra Ferro Freq Con 1994;41:90 –95.
blood vessels. Tortoli P, Guidi G, Newhouse VL. Improved blood velocity estimation
using the maximum Doppler frequency. Ultrasound Med Biol
In summary, we present a convenient flow velocity 1995;21:527–532.
estimation method, which is accurate, and is also easy to
implement in existing Doppler ultrasound systems. One APPENDIX A
additional advantage of this method is that it can measure
Equation (3) minus eqn (1):
the blood flows not only in the longitudinal scan but also
in the cross-sectional scan. This characteristic enables
W
the measurement of curved blood vessels in which the f max ⫺ f d ⫹ 䡠 䡠 sin (A1)
F
longitudinal scan is hard to obtain. Future works will
apply this method in renal arteries, hepatic arteries, aorta,
Equation (A1) divided by eqn (1):
etc.
Acknowledgements—Parts of this work were supported by the National f max ⫺ f d W
⫽ 䡠 tan (A2)
Science Council, R.O.C. fd 2䡠F
REFERENCES where W is the aperture of the transducer, and F is the focal length.
Thus, the beam-vector angle (Doppler angle) is derived:
Andrew JW, Junru W. Correction of intrinsic spectral broadening
errors in Doppler peak velocity measurement made with phased
sector and linear array transducer. Ultrasound Med Biol 1995;21:
1928 –1935.
Bohs LN, Freiemel BH, McDermott BA, Trahey GE. A real time
⫽ tan ⫺1 冋 f max ⫺ f d 2 䡠 F
fd
䡠
W 册 (A3)