Three-dimensional power Doppler imaging: a phantom study to quantify vessel stenosis

Copyright

Ultrawund Cm 1996

in Med. & Biol., Vol. 21, No. X. pp. 1059 - IOh9. I996 World Fedrrstion for Ultrawund in MedIck & Biology Printed in the USA. All right\ reserved 03OI-5629/Yh $7I s.00 + .oo

PII: SO301-5629(96)00125-l

ELSEVIER

*Original

Contribution

THREE-DIMENSIONAL STUDY Imaging

POWER DOPPLER IMAGING: A PHANTOM TO QUANTIFY VESSEL STENOSIS

ZHENYU Guo and AARON FENSTER Research Laboratories, The John P. Robarts Research Institute, London, Ontario, Department of Medical Biophysics, University of Western Ontario, London, Ontario, Canada

Canada;

Abstract-This study investigated whether three-dimensional (3D) power Doppler imaging can be used to quantify arterial stenosis and its potential as an alternative to x-ray angiography. Three-dimensional power Doppler images of in vitro stenotic vessels were generated under different hemodynamic conditions with a 3D power Doppler imaging system. This system includes: a Macintosh Quadra 840AV computer used to perform 3D imaging acquisition, reconstruction and display; a computer-controlled motor-driven translation assembly used to move the transducer; and an ATL Ultramark 9 HDI ultrasound system. Three vascular- and tissue-mimicking phantoms containing three wall-less stenotic vessels with area reduction of SO%, 50% and 30% were imaged with different flow rates under both steady and pulsatile flow conditions and with different Doppler angles under steady flow condition. With the use of the blood mimic, experimental results demonstrated that power Doppler imaging is nearly independent on flow velocity and Doppler angle. It was also demonstrated that 3D power Doppler imaging can produce nonpulsatile angiographic-like 3D images of the flow field. The stenotic vessels were quantified with an overall accuracy of 8.3% of the vessel area and an overall precision of 7% of the vessel area under the conditions described in this paper. It is believed that 3D power Doppler imaging can be used to quantify arterial stenosis, and in some applications it could be an alternative to x-ray angiography. Copyright 0 1996 World Federation for Ultrasound in Medicine & Biology

Key Words: Power Doppler

imaging,

Three-dimensional

imaging,

Doppler

study, we demonstrate the potential use of a new modality of ultrasound imaging, three-dimensional (3D) power Doppler imaging, to depict directly the lumen geometry for stenosisquantification.

INTRODUCTION In colour

Doppler

imaging,

the mean

fre-

quency shift of the local blood flow is encoded in colour and superimposed over the two-dimensional ( 2D) B-mode image (Mitchell 1990). Colour Doppler imaging therefore provides the spatial distribution of flow velocities in relation to the associated anatomic structure. thus depicting the flow field. Unfortunately, the appearance of the flow field in colour Doppler imaging

is highly

dependent

on the instrument

Vessel stenosis.

Maim limitations CJ~’ &our Doppler imaging Some important limitations of mean frequencybasedcolour Doppler imaging have been described by Rubin and Adler ( 1993). One obvious limitation is related to the way that colour Doppler imaging maps the electronic noise as background colour. Because the Doppler frequency shift is dependent on the rate of change of the phase angle, noise with its random phase angle appearsasflow with random velocity and direction, which therefore presentsas random colour. The random colour noise is of the samemagnitude as the true colour from blood flow. Often the flow signal is overwhelmed if the instrument settingsare not properly selected,making the visualisation of the flow field difficult. Other limitations of colour Doppler imaging include Doppler

settings

selected by the examiner (Baumgartner et al. 1991; Guo et al. 1995 ) . In the diagnosis of arterial stenosis, colour Doppler imaging is therefore used mainly for qualitative analysis of blood flow and for displaying the flow field to guide the Doppler sample volume placement for Doppler spectral analysis. In this irz vitro> Address correspondence to: Aaron Fenster, Imaging Research Laboratories, The John P. Robarts Research Institute, P.O. Box 5015, 100 Perth Drive, London. Ontario N6A 5K8. Canada. E-mail: afenster @irus.rri.uwo.ca 1059

Llltraaotmd

in Medicine

and

Biology

angle dependence and tiequency aliasing. Because the Doppler frequency shift is a function of the angle between the ultrasound beam and the direction of motion of the reflectors, flow mapping will have a different appearance with the use of different Doppler angles. In addition. the frequency aliasing distorts the direction and velocity information. making vessels look discontinuous.

Technique o_f’power Doppler imuging As described by Rubin and Adler ( 1993), power Doppler imaging has many promising properties for flow field mapping. It was initially developed for the assessment of low velocity flow (Jain et al. 1991). Instead of displaying the mean Doppler frequency shift, the colour map of power Doppler imaging displays the integrated power of the Doppler signal. The background noise is normally white and uniformly distributed and has lower power than the flow signal, therefore the noise will be represented as a homogeneous background colour even when the gain is increased beyond the level at which the random colour noise dominates the conventional colour Doppler image. This increases the sensitivity of flow mapping. Because power Doppler imaging maps the power of the Doppler signal. it is in some way velocity-independent. Bascom et al. ( 1988) found no change in the power of the Doppler signal scattered by nonaggregating red cells when the mean flow velocity varied between 68 and 100 cm/s. Using a pulsatile flow loop model with polystyrene microspheres as scatterers, Cloutier and Shung ( 1993 ) also observed no significant variation of Doppler power with variation of flow velocity between 10 and 90 cm/s (mean velocity 1 1 cm/s 1. These studies demonstrate that, when there is no red cell aggregation or when a blood mimic is used, power Doppler imaging is velocity-independent. In fact, the Doppler power spectral curve has a wider bandwidth with lower amplitude for high-velocity flow and a narrower bandwidth with higher amplitude for low-velocity flow, making the integral under the curve (power) almost identical. The Doppler signal power is related to the number of moving scatterers (Shung et al. 1976). A change in the Doppler angle does not change the number of scatterers in the sample volume, therefore power Doppler imaging is expected to be Doppler angle-independent. In fact, power Doppler imaging can display flow perpendicular to the ultrasound beam due to the intrinsic spectral broadening. Furthermore, power Doppler imaging is aliasing-free because the integral under the spectral curve is not affected by signal wrapping around PRF/2. This also means that power Doppler imaging should be independent of the PRF setting,

or.. a fixed PRF can be used to image arterie with different flow velocities. Power Doppler imagin, (r has been LIMXI to detect soft-tissue hyperemia (Newman et al. 199~ ) and to depict intrarenal vasculature (Bude et al. 1994 J. In our laboratory. we have been using 3D power Doppler imaging to depict the vasculature of the prostate. liver. kidney and spleen, as well as the flow field of the carotid artery ( Downey and Fenster 1995a, 199% 1. The appearance of the 3D power Doppler images, after they have been rendered in 3D using maximum intrnsity projection (Robb 1995 ). was found to be similar to subtraction angiograms. Three-dimensional imuging ,for arterial stenosis gumtiJication und research objectives Currently, Doppler sonography (with colour Doppler imaging for sample volume placement) is the most widely applied technique for arterial stenosis quantification. However, recent studies have shown that there are limitations in the use of the Doppler sonogram to quantify multisegmental arterial stenoses (Allard et al. 1994). Magnetic resonance angiography ( MRA) can produce images of the arterial lumen, but there are fundamental problems involving signal loss associated with the flow turbulence, making routine assessment of arterial stenosis difficult (Edelman 1992). Because power Doppler imaging has the potential to produce angiographic-like images of the flow lumen without the turbulence-related signal loss seen in MRA. it may be used to quantify arterial stenosis. Conventional x-ray angiography and the current commercially available power Doppler imaging system can only provide images in 2D. If only one plane is imaged, vessel stenoses can be misgraded or even entirely missed. A vessel stenosis may be seen well in one plane and may not be seen at all in others. As a consequence, the stenosis may be more (or less) severe than it appears. Although bi-plane angiography is commonly used to alleviate some of the problems with single-plane imaging, further improvement can be achieved with true 3D imaging. Three-dimensional reconstruction has been applied to B-mode (Belohlavek et al. 1993; Rankin et al. 1993; Roelandt et al. 1994) and colour Doppler (Picot et al. 1991. 1993; Pretorius et al. 1992), and in this study we apply it to power Doppler imaging. Viewing the vascular information in 3D allows the radiologist to rotate and slice the image to reveal a view of any arbitrary plane, which would help in diagnosing arterial stenosis and in following the disease. The nature of power Doppler imaging and the advantages of 3D reconstruction motivated us to investigate whether 3D power Doppler imaging can be used to quantify

Three-dimensional power Doppler imaging 0 Z. Gcro and A. FEKSTER

the degree of arterial stenosisand to study its potential as an alternative to x-ray angiography. As the first step, we validated the technique by using in vitro stenotic vessels.It was the purpose of this study to: ( 1) generate 3D power Doppler images of in vitro stenotic vessels, ( 2) test the dependence of 3D power Doppler imaging on the flow velocity and Doppler angle and (3 ) determine the accuracy and precision of vessel stenosis quantification done by 3D power Doppler imaging.

MATERIALS

AND METHODS

Three-dimensional power Doppler imaging system Figure 1 is a block diagram of our experimental system. For this study, we usedan ATL Ultramark 9 HDI ultrasound system. This system has two colour modes: velocity/variance and power. A 38-mm-aperture highresolution linear array transducer was used, which was operated at 5 MHz for B-mode imaging and 4 MHz for power Doppler imaging. The 3D power Doppler imaging system, developed in our laboratory, includes a Macintosh Quadra 840AV computer and a motor-driven translation assemblycontrolled by the computer (Picot et al. 1993). The computer is capable of image acquisition, reconstruction and display of 3D images in surface rendering and texture mapping modes.The linear array ultrasoundtransducer was mounted on the translation assembly with the transducer axis at a known angle to the imaging plane. A setof planar 2D power Doppler images, transverse to the vessel, was acquired while translating the transducer along the vessel at a fixed Doppler angle. These 2D images were digitised with a precision of 8

1061

bits/pixel by a video-capture board installed in the computer. After acquisition, the images were reformatted into a 3D volume image to compensatefor the Doppler angle and recover the correct geometry. The resulting 3D volume image was viewed on the same computer using 3D viewing software also developed in our laboratory (Fenster et al. 1995). The viewing software allowed rotation of the 3D image and interactive slicing into the volume in any plane in real-time with a computer mouse. Stenotic vessel phantom and blood-mimicking jluid In the present study, the blood vessel phantom was composed of a Plexiglas box containing a tissue mimic and a stenotic wall-less vessel that simulated vessels such as the femoral or the common carotid artery (Rickey et al. 1995 ) . This box was covered by a 0.5-mm-thick urethane layer, which has an ultrasound impedance similar to that of tissue and low attenuation, making it a good acoustic window and preventing the tissue mimic from drying. The tissue mimic (Rickey et al. 1995) was based on a water and agar (high strength A-6924, Sigma Chemical Co.) gel. Glycerol (8%) was added to increase the acoustic velocity to that of real tissue, and 3% 50-pm cellulose scattering particles (S-5504 Sigmacell, Sigma Chemical Co.) were added to provide an acoustic attenuation of 3.5 dB/cm at 4 MHz. The stenotic wall-less vessel was formed by pouring the molten tissue mimic around two brassrods joined at two tapped ends and then removing the two rods from both sides after the tissue mimic has set. The brassrod has a radius of R0 and the tapped end of the rod has a radius given by (Young and Tsai

1973): R(z) =

Fig. 1. Block diagram of the 3D power Doppler imaging system. A Macintosh Quadra 840AV computer digitises the power Doppler video signals from an Ultramark 9 HDI ultrasound system. This computer is also used to perform 3D volume reconstruction, display and analysis. A linear array transducer is moved by a motor-driven stage. Flow of a blood mimic is generated by a computer-controlled pump.

R. - a[1 + COS(~~Z/Z,,)], 0 5 z 5 z(, R,,,

elsewhere

where R0 = 3.98 mm and z. = 10 mm. In the present study, three cosine-shaped stenosesof 80%, 50% and 30% area reduction were created to simulate clinically severe, moderate and mild stenosis. These stenoses corresponded to a = 1.1, a = 0.583 and a = 0.325, respectively. Both sidesof each stenosisextended with radius R0 for a distance greater than the inlet distance. Since the tissue mimic does not absorb water, phantoms were kept in a water bath. The blood-mimicking fluid was a solution of machinist’s cutting fluid ( Acra Tech Syn-Cut HD, Mississauga, Ontario, Canada) and distilled water. The ratio of cutting fluid to water was 60% to 40%. Pulverised nylon 6-12 particles (ELF Atochem Orgasol 3501

1062

Lkasound

in Medicine

and

Biology

EXD ), which have a mean diameter of 10 pm. were added at 15% hematocrit to provide the scattering sites. Using a capillary viscometer (Canon 30), the viscosity of the blood mimic was measured and found to be 0.0206 cm’/s. The acoustic velocity of the blood mimic was 1560 m/s ( Rickey et al. 1995). This working fluid was pumped through our wall-less vessels by using a computer-controlled positive-displacement pump developed in our laboratory (Holdsworth et al. 1991: UHDC flow system, Quest Image Inc., London, Ontario, Canada), as shown in Fig. 1. The pump can produce very accurate steady flow and various pulsatile flows. Instrument

settings

In the Ultramark 9 HDI system, the signal power due to moving scatterers is calculated and colourmapped using a predefined colour map. Background noise is displayed as the low colour value (closest to the colour baseline) and the scatterer motion is displayed as the high colour values (far from the colour baseline). In the power mode, the B-mode image (tissue data) in the colour overlay window was turned off, making the power Doppler image look like a subtraction angiogram. In the present study, a gray-scale was selected to map the Doppler power. Thus, the background noise was displayed as almost black (low power) and the flow field displayed as white (see Fig. 3)) which makes the power Doppler image appear more like a subtraction angiogram. Another advantage of this colour map choice is that we could use a grayscale video-capture board to digitise and process only the gray-scale image. In the ultrasound system, the parameter colour vs. echo write priority (CEWP) is used to select a threshold at which the tissue’s echo data will overwrite the flow data. Above this threshold, echoes are considered to be strong enough to reject flow data, and only tissue data are displayed. Since the tissue data in the coiour overlay window were shut off in the present study, the CEWP was turned off to maximise the sensitivity to the flow. Thus, signals in the power Doppler image are mapped either as flow or as background noise. The persistence level of the system, which determines the number of video frames to be averaged in a weighted manner, can be set between 0 and 7, with a setting of 0 corresponding to no averaging. Because power Doppler imaging is not pulsatile (Bude et al. 1994), a persistence level of 7 was used for our experiments to obtain better flow/no-flow contrast and better vessel contour depiction. The geometry of our stenotic vessel was well defined and provided a reference for determining the colour gain and ultrasound output power of the ultrasound

Volume

2.

h’umhcr

X, IYYh

system. These parameters were selected before the tissue data were shut off in order to optimist the how field visually. Specitically, a colour g;nn of’ 75% and an SPTAd of I5 were chosen. SPTAtl represents the measured value of derated spatial peak time average intensity. which is directly related to the ultrasound power. Other baseline instrument settings included I .5 kHz PRF, 50 Hz wall filter. and I6 for colour sensitiv ity. These instrument settings were kept constant during the study. Imaging

procedures

Three-dimensional power Doppler images were generated for each stenotic vessel with a seriesof flow rates under steady and pulsatile flow conditions as well as with different Doppler angles under steady flow conditions. At first, seven 3D power Doppler images for each stenotic vesselwere generated by using steady flow with seven different how rates between 6 mL/s (Re = 460, mean velocity = 4.2 cm/s) and I8 mL/s (Re = 1400, mean velocity = 12.6 cm/s) in 2-mL/s increments to test the influence of flow velocity on the power Doppler image. The Doppler angle was set to 70” and the coupling fluid between the transducer and the phantom was water. A slow image acquisition rate of 1.5 frames/s was used; however, this rate could be increased if a lower persistence level was used. For each flow rate. 185 contiguous 2D power Doppler images, 0.2 mm apart (over a 37-mm distance ), were acquired for each vessel, with every 2D image having dimensions of I40 X 180 pixels. The acquired 3D power Doppler image was thus a 140 X 180 X 185voxel volume. Working in zoomed mode of the ultrasound system, the resulting voxel dimensions in the .Y. .v and ; directions were calibrated and found to be 0.1078 mm x 0.0835 mm X 0.2 mm. Therefore, the 3D volume had dimensions of 15 mm X 15 mm X 37 mm. Secondly, the above procedure was repeated but using pulsatile flow, which had a velocity waveform similar to that found in the normal human femoral artery (Holdsworth et al. I99 1 ) . Peak flow rates of the pulsatile flow varied between 6 and 18 mL/s in 2-mL/ s increments. A pulsation of 70 beats/min was used for all the pulsatile flow imaging. In the pulsatile flow study, no attempt was made to obtain 3D images at a specific phase of the flow cycle. Finally, 3D power Doppler images of each vessel were generated under a steady flow of 10 mL/s at six different Doppler angles, 60”. 65”, 70”. 75”, 80” and 85”, to test the influence of Doppler angle on the power Doppler image. The 3D power Doppler imageswere displayed with surfacerendering ( as in Fig. 2) and texture mapping (as in Fig. 3) to demonstratethe appearanceof 3D power Doppler imagesof the stenotic vessels.The variation in

Three-dimensional

power Doppler imaging 0 Z. Guo and A. FENSTEK

1063

the appearanceof the flow lumen with varying flow velocity and Doppler angle can be seenin texture-mapped 3D images (as shown in Figs. 8, 9). Analysis To quantify the flow field, the flow area of every 2D slice perpendicular to the vessel axis in every 3D image was calculated by counting pixels corresponding to flow. Because the power Doppler images included only the flow and background noise, a threshold of 10% of the maximum value of the power map was used to separate flow from background noise. The calculated value of flow areawas plotted as a function of its location along the vessel axis to create a flow area curve. This how areacurve was then comparedwith the corresponding true flow area curve by computing the root mean square(RMS) error between thesetwo curves. The FWS errors for different flow rates were compared (Fig. 7a) to investigate the dependenceof power Doppler images on the flow rate (under both steady and pulsatile flow), while the RMS errors for different Doppler angles were compared (Fig. 7b) to investigate the dependence of power Doppler images on the Doppler angle. For each vessel, two mean area curves, one for steady flow and another for pulsatile flow, were obtained by averaging seven curves corresponding to

Fig. 3. Interactive orthogonal displays using texture mapping of the 3D power Doppler imagefrom Fig. 2a. Panelsa, b. c and d correspond to the x, y, z and x&y slicing planes. Using texture mapping, the luminal geometry of the stenotic vessel can be viewed in any orthogonal and oblique plane.

seven flow rates (panels a and b of Figs. 4-6). In addition, one mean area curve was obtained for steady flow by averaging six curves corresponding to six Doppler angles (panels c of Figs. 4-6). The accuracy in quantifying the stenosis was then determined by calculating the residual errors between the mean area curve and the true area curve at each condition. An overall accuracy was calculated as the mean accuracy of all the vesselsunder all the conditions. The precision in quantifying the geometry of each vessel was defined as the fractional standarddeviation (standard deviation divided by the mean) and presented in a figure (Fig. 10). The overall precision was defined as the mean precision of all the vessels under all the conditions.

RESULTS

Fig. 2. Surface renderings of 3D power Doppler images of three stenotic vessels. Panels a, b and c correspond to 80%. 50% and 30% area reduction stenosis, respectively. These 3D images were generated with a peak flow rate of 12 r&/s under pulsatile flow and a Doppler angle of 70”.

Three-dimensional power Doppler imaging Figure 2 shows examples of surface-rendered 3D power Doppler images of the three wall-less stenotic vessels. These 3D images were generated with pulsatile flow having a peak flow rate of 12 mL/s. For surface rendering, the vessel was extracted from the 3D power Doppler volume and depicted by creating surfacesat the boundariesof the vessel. Surfaces facing the observer were displayed, and shadowing techniques were used to provide 3D perspective. Another technique to display 3D data is texture mapping. Figure 3 shows the same data as in Fig. 2a, but here the 3D image has been sliced from different directions and the appropriate 2D image has been texture mapped on the revealed faces of the polyhedron. Using this technique, the 3D image can be interactively rotated

and sliced in any arbitrary plane (orthogonal and oblique i in real-time to reveal a view on that plane, including views impossible to obtain by using conventional techniques. Figure 3a. b, c and d corresponds to slicing planes of .I-. J. .- and .X&Y. We found that it was easier and faster to use texture mapping to identify stcnosesand mimic the appearanceof subtraction angio&rams. It was noticed that, although a pulsatile tlow was used, the appearanceof the power Doppler image was almost uniform. From Figs. 2 and 3, it is clear that the power Doppler image provides stable and continuous anatomical information on the vessel lumen.

-

TrueArea

I

01

-15

MeasuredArea

l

I

I

I

-5 DistAce

-10

it?

I

mm

I

10

15

I

I

20

(4 I

8o -

I

I

-TrueArea

I

MeasuredArea

l

-

Figures 3. 5 and 6 show the measuredmean area curves ((only 70 of the I85 values are displayed here

for each measured curve for clearer viewing) plotted together with the true area curves for the three stenotic vessels. Panels a are mean area curves of seven flow rates under the steady flow condition, panels b are mean area curves of seven peak flow rates under the pulsatile flow condition and panels c are mean area curves of six Doppler angles under the steady flow condition. It can be seen that the measuredflow areas correspond well to the true flow areas, indicating that the 31) power Doppler imagescan be usedto determine vessel stenosis. The similarity of panels a. b and c indicates that similar power Doppler images could be generated in spite of using steady or pulsatile flow, using different flow rates. or using different Doppler angles. This is due to the nature of power Doppler imaging and due to the high persistence level used. The RMS errors between individual areacurves and correspondingtrue area curves for different stenotic vesselsunder different conditions are shown in Fig. 7. Figure 7a show4 the RMS errors at different flow rates under steady and pulsatile flow, while Fig. 7b showsthe RMS errors at different Doppler angles.From the,cefigures, it is apparent that there are small increasesin RMS errors with increasing flow rate. For the 80% stenotic vessel, increasingthe Doppler angleresultedin increasesin RMS errors. However, the differences of these RMS errors are about 10% and are largely due to effects related to the wall filter. asdiscussedlater. From theseresults,we have demonstratedthat, when a blood mimic is used, power Doppler imaging is nearly velocity-independent and nearly Doppler angle-independentunder both steady and pulsatile flow. The near velocity independence is also demonstrated visually in Fig. 8 and the near Doppler angle independence in Fig. 9, since these 3D power Doppler images are quite similar even with different flow rates or different Doppler angles. In addition, effects of pul-

I -10

0 -15

I

I

-5 Didice

I

I

I

in5

10

15

I

I

mm

20

(W

80 -

I -

I

I

TrueArea

I

MeasuredArea

l

-

“E

E 6o

= I 2

40 20

0 I

-15

I

-10

I

I

-5 Di&ce

I

i,”

mm

I

I

10

15

1

20

(4 Fig. 4. Measuredflow areas(mean i I SD) and true flow areasasa function of location along the vessel(called flow areacurvesin the text) for the 80% stenosisphantom.Panel a was computedover seven 3D volumes correspondingto sevenflow rates(range6- 18 mL/s) understeadyflow conditions. Panelb wascomputedover seven3D volumescorrespondingto the sevenpeak flow rates(range 6-18 mL/s) underpulsatileflow conditions.Panelc wascomputedover six 3D volumescorrespondingto six Doppler angles(range 60”-85”) under IO-mL/s steadyflow conditions. satility are not evident in the power Doppler image shown in Fig. 3. which was generated with pulsatile flow without flow cycle gating. Accuracy arzd precision of stenosismeasurement The maximum and mean residual errors between mean area curves and corresponding true area curves,

Three-dimensional

power Doppler imaging 0 2. Guo and A.

shown in Figs. 4-6. are presented in Table I. For example, a mean error of 1.13 mm* (2% of vessel area) and a maximum error of 5.34 mmr (10% of vessel area) were obtained for the 80% stenotic vessel under the pulsatile flow condition. For all vessels and all conditions, the overall maximum error was 7.92 mm2 ( 16% of vessel area) and the overall mean error was 4.15 mm’ ( 8.3% of vessel area). All the errors are positive, indicating overestimation of the flow field, which could be due to the effect of the finite sample volume of the ultrasound system. The mean values of the fractional standard deviations of flow areasfor ev-

O-15

TrueArea

MeasuredArea

l

I

I

-10

-5

I

I

DistArce

in5 mm

I

I

10

1.5

I

I

20

(4 I

80 80 b- -

1065

FENSTER

-

I

I

TrueArea

I

MeasuredArea

l

“E E 60

4

*E 40 I 4 20 0 I -10

01 -15

I I -5 IX&L

I in5

mm

I

I

10

15

I

I

-10

-5

I

1

I

DistAce

it? mm

I

IO

I

15

20

20 (W

(4

80

- -

TrueArea

I NE 80 - -TrueArea E .-c 60

MeasuredArea

l

“E E 60

t 40 2 ii 20 rl: 0

E 40 8 2 20

3 I

0

5

-10

I

I

I

-5 DistAce

in5

mm

I

I

10

15

I

I

-10

-5

80

-

I

O-15

I

DistaOnce i,” mm

I

I

IO

15

-

20

(4

Fig. 6. Measuredflow areas(mean + 1 SD) and true flow areasas a function of location alongthe vesselfor the 30%

I

TrueArea

I

I I I MeasuredArea

20

stenosis phantom. I

I 6

I

(b)

I

1

I

I

See caption of Fig. 4 for explanation.

I

MeasuredArea

l

ery vessel at every condition are presented in Fig. 10. The overall mean value is 7%. Stenoses can thus be determined with an overall accuracy of 8.3% of the vessel area and an overall precision of 7% of the vessel area under the conditions described in this paper. I

I

-10

-5

I

DiotcfL

I

in5 mm

I

I

IO

15

J

20

(4 Fig. 5. Measured flow areas (mean 2 1 SD) and true flow areas as a function of location along the vessel for the 50% stenosis phantom. See caption of Fig. 4 for explanation.

DISCUSSION

AND CONCLUSION

Because cardiovascular physiology is essentially 3D, conventional 2D techniques provide only a small fraction of the total information, requiring diagnosticians to perform mental reconstructions to appreciate

Llltrasound in Medicine and Biology

Volume 22. Number X. 1996

colour Doppler image to localiae functional change\ relative to the underlying anatomy. Because power Doppler imaging maps the flow field. the .iD power Doppler image hasthe potential to be useful in qualitative and quantitative assessmentof both the geometry of arterial lumen and pathologic aspects of the artery. The variances of the luminal areas seen in Figs. 4-6 could be due to the effect of the wall tiltrr. Normally, an increase of the wall filter decreasesthe Row area (Jain et al. 1991) and vice versa. In the present study, a constant wall filter ( 50 Hz) was used, having a larger effect on low-velocity flow and smaller effect on high-velocity flow. Figures 4-6 show that the variances are larger in areas distal and proximal to the stenosis, where low-velocity regions are present. and much smaller at the neck of the stenosis. where the flow velocity is high. With the rigid vessels used in the present in vitro study, if the wall tilter could be

Flow Rate in ml/s (4 0.4

0.3 2 0 t w 0.2 i 0.1

0 55

60

65

Doppler

70

Angle

75

80

85

90

in Degree

(W Fig. 7. (a) RMS errorsbetweenthe individual measuredarea curve and the true area curve as a function of flow rate. VsOS,80% stenosiswith steady flow; VsOP,80% stenosis with pulsatile flow; V&S, 50% stenosiswith steady flow; V5”P, 50% stenosiswith pulsatileflow; V&S, 30% stenosis with steady flow; V3,,P, 30% stenosiswith pulsatile flow. The flow rate representsthe meanflow rate for steadyflow and the peak flow rate for pulsatile flow. (b) RMS errors between the individual measuredarea curve and the true area curve as a function of Doppler angle. VaO,VzO and V3,,correspondto the 80%, 50% and 30% stenotic vessels, respectively. the 3D structure and pathologic condition of the vasculature. Rankin et al. ( 1993) reported that 3D ultrasound images can clearly display anatomic details and spatial relations and can be manipulated interactively to improve the clinician’s visualisation. The results of Picot et al. ( 1991, 1993) demonstrated the ability of the 3D

Fig. 8. Longitudinal slicing planesof 3D power Doppler imagesof the 50% stenoticvesselwith steadyflow. Panels a, b, c and d are for flow rates of 6, 10. 14 and 18 mL/s, respectively.

Smaller vessel diameter can be seen in panel a

becausethe low-velocity flow nearthe wall waseliminated by the wall filter.

Three-dimensional

power Doppler imaging 0 Z. Guo and A.

FENSTER

1067

0.2

1

0.15 c 2 z

0.1

e 0.05 0 1

IAl

3

Fig. 10. Mean values of fractional standard deviations of flow areas. S, steady flow; P, pulsatile flow; A, angle. Vessels 1, 2 and 3 correspond to the 80%, 50% and 30% stenotic vessels.

Fig. 9. Longitudinal slicing planes of 3D power Doppler images of the 80% stenotic vessel with steady flow of 10 mL/s. Panels a, b and c correspond to Doppler angles of 60”, 70” and 80”, respectively.

turned off, much smaller area variances could be expected along the vessel; however, this would not be appropriate when imaging humans due to the pulsation of the vessel wall. As mentioned in the introduction, when there is no red cell aggregation or when using a blood mimic asthe scattering fluid, power Doppler imaging is nearly velocity-independent. It has been recognised that in the presence of flow turbulence and red cell aggregation, Doppler backscattered power from blood will be increased(Cloutier et al. 1995) . Becauseflow turbulence and red cell aggregation are related to the flow velocity, the variation of flow velocity will change the pixel intensity of the power Doppler image. However, if

power Doppler imaging is used merely to determine whether blood flow is present or not (the angiographic application), variations of image pixel brightness do not change the estimate of flow fields. Furthermore, these variations could be reduced by increasing the colour gain and ultrasound output power of the system. With the use of the blood mimic in the present study, it was found that there is no significant change in the power Doppler images upon varying the flow velocity. As discussed earlier, the power Doppler image should be independent of the Doppler angle. In practice, however, the image manifests slight angle dependence. In our study, with an increase of the Doppler angle, the neck of the stenosisbecame longer in the 3D image (Fig. 9). This occurred because in flow separation regions, immediately upstream and downstream of the stenosisneck, the flow velocities are low and the imaging system with a given wall filter setting cannot display them when a large Doppler angle is used. If the wall filter could be turned off, power Doppler images should be angle-independent. In fact, the power Doppler imaging system can display flow even when the mean Doppler frequency shift is zero, as a result of spectral broadening produced by the aperture of the scan head (Bude et al. 1994). However. in

Table 1. Maximum and mean errors between measured flow area and true flow area for different vessels under different conditions. Vessels, 1, 2 and 3 correspond to 80%, 50% and 30% stenosis. Vessel 1: Error (mm*)

Vessel 2: Error (mm’)

Vessel 3: Error (mm’)

Varying parameter

Max.

Mean

Max.

Mean

Max.

Mean

Velocity (steady) Velocity (pulsatile) Doppler angle

8.55 5.34 7.61

5.56 1.13 3.51

9.58 8.13 9.61

6.31 4.9 1 3.91

7.64

3.7-t 3.3 3.95

5.94 8.87

106s

Ultrasound

in Medicine

and Biology

clinical practice. the wall filter is necessary to minimise soft-tissue and vessel wall motion; otherwise, these motions can seriously degrade the images. Even with this slight angle-dependence associated with the wall tilter, power Doppler imaging is much less angle-dependent than colour Doppler imaging, and thus is more robust (Rubin et al. 1994 1 in mapping the flow field. It should be noted that the velocity waveform of the pulsatile flow used in this study was tri-phased with two zero-crossings. The velocities near zero were eliminated by the wall filter, which should result in fuctuations in the image. However. these fluctuations were minimised by using a high persistence level, and the power Doppler images in this study appeared almost constant and not pulsatile. The presence of multiple stenoses decreases the accuracy of Doppler sonography for stenosis quantification. However, multiple stenoses should not affect power Doppler imaging. Magnetic resonance angiography can now produce 3D images of arterial lumens, but its ability for stenosis quantification is limited due to the inherent problem of signal loss, as mentioned in the introduction. The preliminary results of this study show that 3D power Doppler imaging is a promising technique for generating 3D angiographic-like images of arteries and for quantifying arterial stenoses. Further study is required to compare this technique with Doppler sonography and 2D power Doppler imaging. In addition, this 3D technique must be evaluated in patients, where calcification may prevent collection of adequate power Doppler images. Although Doppler sonography has been used in many medical centres to replace xray angiography for diagnosis of arterial stenosis, a 3D morphologic description of the arterial stenosis is very helpful in planning surgery. The technique proposed in this study could be promising for this purpose.

SUMMARY In this study, we generated 3D power Doppler images of in vitro stenotic vessels. We tested the effect of flow rate and Doppler angle on the 3D power Doppler images and demonstrated that, with the use of a blood mimic, the power Doppler image is nearly flow rate- and Doppler angle-independent. The in vitro stenotic vessels tested in this study were quantified with an overall accuracy of 8.3% of the vessel area and an overall precision of 7% of the vessel area under the conditions described. From this in vitro study. it is believed that 3D power Doppler imaging has the potential to be useful in quantifying arterial stenosis and could be an alternative to xray angiography in some applications. Acknowledgetnerzts-We Medical Research

Council

are grateful for the financial support of the of Canada. The first author gratefully

Volume

22. Number-

X. 1996

acknowledges a postdoctoral fellowship ot FLAK I Fontis pour Ia Formation de Chercheurs et L’aide a la Recherche of Quebec ) Wc thank Advanced Technology Laboratories for providing the I’ltramark 9 HDI used in the study. We al$o thank Lori Gard. Sham Dunne, Jan Larson. Bjarne Hansen. Thomas Chan. Shidong Tong, Paul Picot and John Miller for their conrribution~ rn the 3D ultrasound imaging project

REFERENCES Allard L, Cloutier G. Durand L-G. Roederer GO, Langlois YE. Limitations of ultrasonic duplex scanning for diagnosing low’er limb arterial stenoses in presence of adjacent segment disease. J Vast Surg 1994: 19: 650-657. Bascom PAJ, Routh HF. Cobbold RSC. interpretation of power changes in Doppler signals from human blood-br vitro studies. Proceedings IEEE Ultrasonics Symposium, October I988:985-988. Baumgartner H, Schima H, Kuhn P. Importance of technical variables for quantitative measurements by colour Doppler imaging. Am J Cardiol 1991;67:314-315. Belohlavek M. Foley DA. Gerber TC. Greenleaf JE. Seward JB. Three-dimensional ultrasound imaging of the atria1 septum: Normal and pathologic anatomy. J Am Co11 Cardiol 1993; 22: I6731678. Bude RB, Rubin JM, Adler RS. Power Doppler versus conventional colour Doppler sonography: Comparison in the depiction of normal intrarenal vasculature. Radiology 1994; 192:777-780. Cloutier G, Allard L, Durand L-G. Changes in ultrasonic Doppler backscattered power downstream of concentric and eccentric stenoses under pulsatile flow. Ultrasound Med Biol 1995:21 :S9-70. Cloutier G. Shung KK. Cyclic variation of the power of ultrasonic Doppler signals backscattered by polystyrene microspheres and porcine erythrocyte suspensions. IEEE Tram Biomed Eng 1993:40:953-962. Downey DB, Fenster A. Vascular imaging with a three-dimcn,ional power Doppler system. AJR 1995a: 16.5 (3 I :665-668. Downey DB, Fenster A. Three-dimensional power Doppler detection of prostatic cancer. AJR 1995b: 165:741. Edelman RR. Basic principles of magnetic resonance angiography. Cardiovasc Intervent Radio1 1992; 15:3- 13. Fenster A. Dunne S. Chan T. Method and system for constructing and display three-dimensional images. US Patent, no. 5454371. 3 October 1995. Guo Z, Moreau M. Rickey DW. Picot PA. Fenster A. Quantirativr investigation of in vitro flow using three-dimensional colour Doppler ultrasound. Ultrasound Med Biol 1995; 2 1:807-8 16. Holdsworth DW, Rickey DW. Drang,ova M. Miller DJM, Fenstet A. Computer-controlled positive dtsplacement pump for physiological flow simulation. Med Biol Eng Comput 1991:29:565570. Jain SP. Fan PH, Philpot EF, Nanda NC, Aggarwal KK. Moos S. Yoganathan AP. Influence of various instrument settings on the flow information derived from the power mode. Ultrasound Med Biol 1991: 17:49-54. Mitchell DG. Colour Doppler imaging: Principles. limitations, and artifacts. Radiology 1990; 177: I - IO. Newman JS, Adler RS, Bude RO. Rubin JM. Detection of softtissue hyperemia: Value of power Doppler sonography. AJR 1994; 163:385 -389. Picot PA, Rickey DW. Mitchell R, Rankin RN. Fenster A. Threedimensional colour Doppler imaging of the carotid artery. In: Kim Y, ed. Image capture, formatting, and display. SPIE Proceedings. volume 1444. Bellingham, WA: SPIE-The International Society for Optical Engineering, 1991:206-2 1.1. Picot PA, Rickey DW. Mitchell R. Rankin RN, Fenster A. Threedimensional colour Doppler imaging. Ultrasound Med Biol 1993; 19:95- 103. Pretorius DH. Nelson TR, Jaffe JS. Three-dimensional sonographic analysis based on colour flow Doppler and gray scale image data: A preliminary report. J Ultrasound Med 1992; 11:225-232.

Three-dimensional

power

Doppler

Kankin RN, Fenster A. Downey DB, Munk PL, Levin MF, Vellet AD. Three-dimensional sonographic reconstruction: Techniques and diagnostic applications. AJR 1993; 161:695-702. Rickey DW, Picot PA, Christopher DA. Fenster A. A wall-less vessel phantom for Doppler ultrasound studies. Ultrasound Med Biol 199S;21:1163-I 176. Robb R. Three-dimensional biomedical imaging. New York: VCH Publishers: 1995. Roelandt TC Jr. Cate FJ, Veltter WB, Taams MA. Ultrasonic dynamic three-dimensional visualisation of the heart with a multi-

imaging

l 2. Guo

and A. FENSTER

1069

plane transesophageal imaging transducer. J Am Sot Echocardiogr 1994:7:217-229. Rubin JM. Adler RS Power Doppler expands standard colour capability Diagn Imaging December 1993 :66-69. Rubin JM, Bude RO, Carson PL. Bree RL, Adler RS. Power Doppler US: A potentially useful alternative to mean frequency-based colour Doppler US. Radio1 1994; 190:853-856. Shung KK, Sigelmann RA, Reid JM. Scattering of ultrasound by blood. IEEE Trans Biomed Eng 1976:23:460-467. Young DF, Tsai FY. Flow characteristics in models of arterial stenoses-I. Steady flow. J Biomech 1973:6:395-410.