Upload
naohiko-tanaka
View
214
Download
2
Embed Size (px)
Citation preview
RAPID COMMUNICATION
Experimental verification of color flow imaging basedon wideband Doppler method
Naohiko Tanaka
Received: 11 February 2013 / Accepted: 22 April 2013
� The Japan Society of Ultrasonics in Medicine 2013
Abstract The purpose of this study is to eliminate the
aliasing in color flow imaging. The wideband Doppler
method is applied to generate a color flow image, and the
validity of the method is experimentally confirmed. The
single beam experiment is carried out to confirm the
velocity estimation based on the wideband Doppler
method. The echo data for the conventional pulsed Doppler
method and the wideband Doppler method are obtained
using a flow model, and the estimated velocity for each
method is compared. The color flow images for each
method are also generated using several types of flow
model. The generated images are compared, and the
characteristics of the imaging based on the wideband
Doppler method are discussed. The high velocity beyond
the Nyquist limit is successfully estimated by the wideband
Doppler method, and the availability in low velocity esti-
mation is also confirmed. The aliasing in color flow images
is eliminated, and the generated images show the signifi-
cance of the elimination of the aliasing in the flow imaging.
The aliasing in color flow imaging can be eliminated by the
wideband Doppler method. This technique is useful for the
exact understanding of blood flow dynamics.
Keywords Color flow imaging � Pulsed Doppler �Aliasing � Nyquist limit
Introduction
Aliasing is one of the major problems in color flow
imaging for the visualization of blood flow. The blood
flow, the velocity of which exceeds the Nyquist limit,
will be observed as a flow of opposite direction because
of the aliasing. This phenomenon has been explained by
the sampling theorem. The target velocity is estimated
by the phase changes of echoes in the conventional
pulsed Doppler method applied to the medical color
flow imaging system. The phase information can be
obtained as a discrete time signal sampled at pulse
repetition frequency, because one phase data is detected
by one pulse transmission. Thus, the velocity, which can
be correctly estimated, is limited by the sampling
theorem.
In order to overcome the aliasing problem, various
methods have been investigated since the basic principle of
color flow imaging was proposed by Kasai et al. [1]. The
use of a transmission pulse with two different frequencies
has been studied [2, 3]. The phase changes are detected for
each frequency component of echoes, and the target
velocity is determined by the difference of each phase
change. The resulting phase difference will correspond to
that which is obtained by the pulse transmission with the
difference frequency. Since the difference frequency can be
made much smaller than each frequency, the velocity range
can be expanded. A similar idea was applied to a non-
equally-spaced pulse transmission method [4]. However,
the use of a second-order phase difference causes a rise in
the influence of noise.
A time-domain correlation technique has also been
studied [5–8]. This scheme is based on estimation of the
time delay of echoes using a cross-correlation function.
The cross-correlation-based methods have a weakness in
N. Tanaka (&)
Department of Electronic Information Systems,
Shibaura Institute of Technology, 307 Fukasaku,
Minuma-ku, Saitama 337-8570, Japan
e-mail: [email protected]
123
J Med Ultrasonics
DOI 10.1007/s10396-013-0462-3
that the influence of the velocity variance of targets is
relatively large.
Elimination of the aliasing in the estimation of blood
velocity could possibly be realized by the wideband
Doppler method [9], which utilizes all the information of
the phase spectrum of echoes obtained by the wideband
pulse transmission. The phase correction in this method
enables estimations of high velocity beyond the Nyquist
limit. Its robustness with respect to noise and the velocity
variance was confirmed by computer simulation. The
principle of the wideband Doppler method and the simu-
lation results were described in a previous paper [9]. Use of
the wideband echo information was also studied by Kondo
et al. [10]. The target velocity is estimated by using only
the gradient of the phase spectrum in this method, resulting
in insufficient robustness.
The purpose of this study is to experimentally val-
idate the performance of the wideband Doppler
method. Several types of flow model are employed for
the acquisition of echo data. The color flow images are
generated by the conventional pulsed Doppler method
and the wideband Doppler method. The resulting ima-
ges yielded by the two methods are compared and
discussed.
Materials and methods
Basic stance of the experimental study
In general, a new method should be verified by using
more reliable data obtained by another method. In the
case of this study, it is ideal to measure the velocity
profile in the flow model by using a non-ultrasound-based
method with higher accuracy. However, it is difficult to
measure the velocity with the same scan format and with
the same spatial resolution at the same time as the
ultrasonic measurement. To avoid this difficulty, the
estimated velocity yielded by the wideband Doppler
method is compared with the result yielded by the con-
ventional pulsed Doppler method.
Experimental setup for the confirmation of velocity
estimation using a single beam transducer
A simple flow phantom and data acquisition system was
developed for confirmation of the velocity estimation of
the wideband Doppler method. The schematic diagram of
this system is shown in Fig. 1. Two fluid bags are con-
nected by a rubber tube with an internal diameter of
4 mm, and are filled with the Doppler test fluid (ATS
Model 707). Bidirectional flow is easily realized by
pressing these bags alternately. The pulser generates the
transmission pulse. A four-cycle burst pulse and one-
cycle burst pulse are used for the conventional pulsed
Doppler and the wideband Doppler measurement,
respectively. A sequence of eight pulses for the conven-
tional Doppler and a sequence of eight pulses for the
wideband Doppler are alternately transmitted to obtain
the echo signal for each Doppler method at almost the
same time. The transmission pulse for the wideband
Doppler measurement has the same amplitude as that in
the conventional Doppler measurement. Because the burst
cycle is reduced, the signal-to-noise ratio (SNR) of the
echo signal for the wideband Doppler measurement is
lower than that in the conventional pulsed Doppler
measurement. A concave PZT (13/, 75R) transducer with
a 75 % fractional band width is used in this experiment.
The parameters of this experiment are shown in Table 1.
The received echo signal is amplified and then A/D
converted with 14-bit resolution at a 50-MHz sampling
rate. The A/D converter used here contains a large-sized
buffer memory, and the echo data obtained during a
10-second period can be stored in the buffer memory.
The digitized echo data are then sent to a PC, and all the
signal processing is carried out offline on the PC. The
Fig. 1 Schematic diagram of the single beam experiment for the
confirmation of the velocity estimation
Table 1 Parameters for the single beam experiment
Parameter Value
Center frequency of transmission pulse 2 MHz
Burst cycle of transmission pulse
Conventional pulsed Doppler 4
Wideband Doppler 1
Number of pulse transmissions (packet size) 8
Pulse repetition time 256 ls
Range gate size 5.12 ls
J Med Ultrasonics
123
audible Doppler signal is also generated for setting of the
transducer position.
Experimental setup for color flow imaging
In order to generate the color flow image of the flow
model, a linear array transducer and a multi-channel data
acquisition system are used as shown in Fig. 2. The array
consists of a 128-element transducer with a center fre-
quency of 7.5 MHz. Table 2 shows the parameters for
the imaging. The 32-ch pulser and the 32-ch A/D con-
verter with 12-bit resolution are connected to the array
transducer through a cross-point switch. The transmission
pulse for the wideband Doppler measurement has the
same amplitude as that in the conventional Doppler
measurement. The beam scanning and the pulser control
in this system are carried out by specified hardware. On
the other hand, receiver beam forming, Doppler velocity
estimation, scan conversion, and all other processing for
generating the color flow image are carried out by soft-
ware on an offline PC. The buffer memory to store echo
data has the capacity for 8-pulse transmission, i.e., the
packet size is 8. Because of some limitations in the
circuit design, a sufficient amount of buffer memory
could not be realized. Consequently, the data acquisition
for the conventional Doppler method and that for the
wideband Doppler method are carried out individually
within a few minutes. During the data acquisition for the
two methods, the flow phantom is carefully controlled to
maintain a uniform condition, and thus the flow models
in the two cases have almost the same velocity profile,
but are not exactly equal. Thus, the images yielded by
the two methods can be compared qualitatively.
Results
Single beam experiment
The results of the velocity estimation obtained by the fixed
ultrasound beam are shown in Figs. 3 and 4. In these cases,
the fluid bags are pressed every few seconds alternately to
generate bidirectional flow. Figure 3 shows the estimation
result of the conventional pulsed Doppler method. The
Nyquist limit in this case is ±0.73 m/s, and is shown by the
dotted line in Fig. 3. The bidirectional velocity can be
identified in Fig. 3, but the estimated velocity aliases at the
Nyquist limit. Figure 4 shows the estimation result of the
wideband Doppler method. The estimation range of
velocity is set to ± 2.05 m/s, and a 35 pattern is employed
for the phase correction. The velocity beyond the Nyquist
limit is also estimated correctly. For the comparison of
these results, the estimated velocity of the wideband
Fig. 2 Schematic diagram of
the color flow imaging
Table 2 Parameters for color flow imaging
Parameter Value
Center frequency of transmission pulse 7.5 MHz
Sampling frequency 50 MHz
Burst cycle of transmission pulse
Conventional pulsed Doppler 4
Wideband Doppler 1
Number of pulse transmissions (packet size) 8
Pulse repetition time 250 ls
Nyquist limit (conventional pulsed Doppler) 0.2 m/s
Ultrasound beam 128 line
Array pitch 0.4 mm
J Med Ultrasonics
123
Doppler method is forcibly aliased and superimposed upon
the result of the conventional Doppler method, as shown in
Fig. 5. The good agreement of these results can be con-
firmed by Fig. 5.
Experiment of color flow imaging
The color flow imaging using the single tube flow model
is carried out for two flow velocities. In all cases, no
image processing is applied, and the color of a pixel is
determined by the estimated velocity of its position.
Figure 6 is generated by using the conventional pulsed
Doppler method. The fluid is moving from the lower left
to the upper right of this figure, so the area in the tube
should be filled by a red color. The blue region of Fig. 6
is caused by aliasing. The result of the wideband Doppler
method is shown in Fig. 7. Although a small number of
error pixels exist, almost the entire region in the tube is
filled by the red color, and aliasing is not observed. The
image pattern of the non-aliased region in Fig. 6 is the
same as that in Fig. 7. This means that the flow is suf-
ficiently stable.
Faster flow is also tested using the same flow model.
The color flow image for a faster flow generated by the
wideband Doppler method is shown in Fig. 8. In this case,
the flow velocity is about twice that in Fig. 7, and the
velocity range in the wideband Doppler method is three
times that of the conventional pulsed Doppler method. A
considerable number of error pixels (blue) are observed in
Fig. 8. The color flow image with reduced spatial resolu-
tion is shown in Fig. 9. This image is generated using the
same data as used in Fig. 8, but the adjacent two scan lines
are combined and the range gate size is twice that of Fig. 8.
The number of the error pixels is decreased by reducing
spatial resolution.
Fig. 3 Estimated velocity by using the conventional pulsed Doppler
method. The dotted lines show the Nyquist limit
Fig. 4 Estimated velocity by using the wideband Doppler method
Fig. 5 Comparison of the estimation result. The red plot is the result
of the wideband Doppler method, and is forcibly aliased. The black
plot is the result of the conventional pulsed Doppler method
Fig. 6 A color flow image of the single tube model. The velocity is
estimated by the conventional pulsed Doppler method
J Med Ultrasonics
123
The advantage of the wideband Doppler method will be
clear in the experiment using the jet flow model. The color
flow images generated by the conventional pulsed Doppler
method and the wideband Doppler method are shown in
Figs. 10 and 11, respectively. The nozzle is placed at the
lower right, and the fluid spurts from there to the upper left
Fig. 7 A color flow image of the single tube model. The velocity is
estimated by the wideband Doppler method
Fig. 8 A color flow image for the faster flow of the single tube
model. The velocity is estimated by the wideband Doppler method
Fig. 9 A color flow image for the faster flow of the single tube
model. The velocity is estimated by the wideband Doppler method.
An enlarged sample volume is used
Fig. 10 A color flow image of the jet flow model. The velocity is
estimated by the conventional pulsed Doppler method
J Med Ultrasonics
123
of the figure. The velocity of the fluid has a maximum
value in the nozzle. Obviously, aliasing is observed in the
image yielded by the conventional pulsed Doppler method
shown in Fig. 10. In contrast, the flow velocity is estimated
successfully in the case of the wideband Doppler method,
as shown in Fig. 11. The slow flow around the jet is also
imaged properly in Fig. 11.
The double tube flow model is employed to confirm the
estimation of low speed flow. Figure 12 shows the image
generated by the conventional pulsed Doppler method. The
fluid in the upper tube is moving from left to right, and the
fluid in the lower tube is moving in the direction opposite
that in the upper tube. The lower flow is faster than the
upper flow. Therefore, the upper and lower tube should be
filled by dark red and light blue, respectively. The red color
in the lower tube in Fig. 12 is caused by the aliasing. The
result of the wideband Doppler method is shown in Fig. 13.
In this case, the flow in each tube is correctly imaged, but it
may be hard to see the dark red in the upper tube. The color
coding on a logarithmic scale is useful for the expression of
wide range velocity. Figure 14 shows the color flow image
on a logarithmic scale. Both the fast flow and the slow flow
are clearly imaged without aliasing.
Discussion
The validity of the velocity estimation based on the
wideband Doppler method is experimentally confirmed by
using several flow models. In all cases, the result generated
Fig. 11 A color flow image of the jet flow model. The velocity is
estimated by the wideband Doppler method
Fig. 12 A color flow image of the double tube model. The velocity is
estimated by the conventional pulsed Doppler method
Fig. 13 A color flow image of the double tube model. The velocity is
estimated by the wideband Doppler method. The upper tube is filled
by dark red because of low velocity
J Med Ultrasonics
123
by the wideband Doppler method is free from aliasing. In
particular, the results of the jet flow model express the
significance of the aliasing elimination in color flow
imaging.
The results of the single beam experiment show that the
estimated velocity in the wideband Doppler method is in
agreement with that in the conventional pulsed Doppler
method. In addition, it is found that good stability in the
estimation of low velocity is achieved by the wideband
Doppler method. The line fitting process in the method will
contribute to the stability in velocity estimation. The
advantage in low velocity estimation can also be confirmed
in Figs. 11 and 14.
As mentioned above, the electric power of the trans-
mission pulse in the wideband Doppler method is lower
than that in the conventional method, so the SNR of the
echo signal for the wideband Doppler method is also lower
than that in the conventional method. The results of the
wideband Doppler method are obtained under the lower
SNR condition. Though the wideband system is, in general,
considered to be sensitive to noise, the proposed method
has good robustness for noise. The line fitting process
contributes to the robustness for noise.
The results of the fast flow imaging shown in Figs. 8 and
9 suggest a relation between the spatial resolution and the
maximum velocity in the estimation. The difference
between Figs. 8 and 9 is in the size of the sample volume
for the velocity estimation. The flow velocity is estimated
by the echoes from the target within the sample volume.
Because the targets are moving, a part of the targets will
enter the sample volume, and a part of the targets in the
sample volume will exit. This causes changes in the echo
waveform and a decline in the correlation with subsequent
echoes. The ratio of replaced targets is determined by the
target velocity and the size of the sample volume. Conse-
quently, the sample volume must be enlarged for faster
flow estimation.
A small number of error pixels exist in the results of the
wideband Doppler method. This error is caused by acci-
dental aliasing [11], which originates from the abnormal
narrowing of the echo spectrum resulting from the inter-
ference of echoes. The wideband Doppler method is based
on the scheme that the target velocity is estimated by using
all the phase information of echoes in many frequencies.
When the bandwidth of the echo spectrum becomes nar-
row, the estimation of the wideband Doppler method will
degenerate into that of the conventional pulsed Doppler
method. There are two ways to suppress the error rate
caused by accidental aliasing. The first way is to make the
bandwidth of the transmission pulse wider, and the other is
to expand the range gate. Image processing, such as using a
median filter, is also effective in making the error pixels
invisible.
Conclusion
The wideband Doppler method for the elimination of ali-
asing in color flow imaging is investigated in this study.
High velocities beyond the Nyquist limit are successfully
estimated by the wideband Doppler method, and the
availability in low velocity estimation is also confirmed.
The color flow images of the flow model show the sig-
nificance of the elimination of aliasing.
The hard limit of the velocity range in the wideband
Doppler method is determined by the number of the phase
correction pattern. In addition, the error rate in the velocity
estimation is determined by the size of the sample volume,
the bandwidth, and the SNR of the echo signal. In blood
flow imaging, the error rate is the dominant factor that
determines the velocity range. Further study, including
statistical analysis of the error rate, will be required.
Quantitative evaluation of the velocity estimation will also
be required in a future study.
The recognition of aliasing in clinical situations is now
dependent upon the skill of the ultrasonographer or medical
doctor. The proposed method will release these profes-
sionals from the burden of recognition of aliasing, and will
contribute to the exact understanding of blood flow
dynamics.
Fig. 14 A color flow image of the double tube model. The velocity is
estimated by the wideband Doppler method. Color coding on a
logarithmic scale is used
J Med Ultrasonics
123
Conflict of interest The author has no conflicts of interest or rela-
tionships to disclose.
References
1. Kasai C, Namekawa K, Koyano A, et al. Real-time two-dimen-
sional blood flow imaging using an autocorrelation technique.
IEEE Trans Sonics Ultrasonics. 1985;SU-32:458–64.
2. Powers JE. Multiple demodulation frequency Doppler. U.S.
Patent 1985; 4 534 357.
3. Nitzpon HJ, Rajaonah JC, Burckhardt CB, et al. A new pulsed
wave Doppler ultrasound system to measure blood velocities
beyond the Nyquist limit. IEEE Trans Ultrason Ferroelectr Freq
Control. 1995;42:265–79.
4. Nishiyama H, Katakura K. Non-equally-spaced pulse transmis-
sion for non-aliasing ultrasonic pulsed Doppler measurement.
J Acoust Soc Jpn (E). 1992;13:215–22.
5. Lai X, Trop H, Kristoffersen K. An extended autocorrelation
method for estimation of blood velocity. IEEE Trans Ultrason
Ferroelectr Freq Control. 1997;44:1332–42.
6. Wu H, Ohtsuki S. An Eulerian segment-matching method for
blood flow velocity estimation using pulsed ultrasound RF ech-
oes. J Acoust Soc Jpn(E). 1998;19:51–8.
7. Hein IA, Chen JT, Jenkins WK, et al. A real-time ultrasound
time-domain correlation blood flowmeter: part I—theory and
design. IEEE Trans Ultrason Ferroelectr Freq Control. 1993;40:
768–75.
8. Hein IA, Chen JT, O’Brien WD Jr. A real-time ultrasound time-
domain correlation blood flowmeter: part II—performance and
experimental verification. IEEE Trans Ultrason Ferroelectr Freq
Control. 1993;40:776–85.
9. Tanaka N. Wideband Doppler method for alias-free estimation of
blood velocity. Jpn J Med Ultrasonics. 2008;35(3):315–28.
10. Kondo M, Kanai H, Chubachi N. A new ultrasonic pulsed
Doppler-based method for measuring higher velocity compo-
nents. T.IEE Jpn. 1995;115-C:1425–31.
11. Tanaka N. Accidental aliasing in wideband Doppler method.
2007 Spring Meet Acoustical Soc Jpn. 2007;1079–80.
J Med Ultrasonics
123