Ultrasound in Med. & Biol., Vol. 40, No. 11, pp. 2700–2706, 2014 Copyright Ó 2014 World Federation for Ultrasound in Medicine & Biology Printed in the USA. All rights reserved 0301-5629/$ - see front matter
http://dx.doi.org/10.1016/j.ultrasmedbio.2014.06.001
d
Original Contribution NOVEL FLOW QUANTIFICATION OF THE CAROTID BULB AND THE COMMON CAROTID ARTERY WITH VECTOR FLOW ULTRASOUND MADS MØLLER PEDERSEN,* MICHAEL JOHANNES PIHL,y PER HAUGAARD,z k KRISTOFFER LINDSKOV HANSEN,* THEIS LANGE,x LARS L€ ONN,* MICHAEL BACHMANN NIELSEN,* and JØRGEN ARENDT JENSENy * Department of Radiology, Rigshospitalet, University of Copenhagen, Copenhagen, Denmark; y Center for Fast Ultrasound Imaging, Department of Electrical Engeneering, Technical University of Denmark, Lyngby, Denmark; z B-K Medical, Herlev, Denmark; x Department of Biostatistics, University of Copenhagen, Copenhagen, Denmark; and k Department of Vascular Surgery, Rigshospitalet, University of Copenhagen, Copenhagen, Denmark (Received 6 December 2013; revised 2 April 2014; in final form 3 June 2014)
Abstract—Abnormal blood flow is usually assessed using spectral Doppler estimation of the peak systolic velocity. The technique, however, only estimates the axial velocity component, and therefore the complexity of blood flow remains hidden in conventional ultrasound examinations. With the vector ultrasound technique transverse oscillation the blood velocities of both the axial and the transverse directions are obtained and the complexity of blood flow can be visualized. The aim of the study was to determine the technical performance and interpretation of vector concentration as a tool for estimation of flow complexity. A secondary aim was to establish accuracy parameters to detect flow changes/patterns in the common carotid artery (CCA) and the carotid bulb (CB). The right carotid bifurcation including the CCA and CB of eight healthy volunteers were scanned in a longitudinal plane with vector flow ultrasound (US) using a commercial vector flow ultrasound scanner (ProFocus, BK Medical, Denmark) with a linear 5 MHz transducer transverse oscillation vector flow software. CCA and CB areas were marked in one cardiac cycle from each volunteer. The complex flow was assessed by medical expert evaluation and by vector concentration calculation. A vortex with complex flow was found in all carotid bulbs, whereas the CCA had mainly laminar flow. The medical experts evaluated the flow to be mainly laminar in the CCA (0.82 ± 0.14) and mainly complex (0.23 ± 0.22) in the CB. Likewise, the estimated vector concentrations in CCA (0.96 ± 0.16) indicated mainly laminar flow and in CB (0.83 ± 0.07) indicated mainly turbulence. Both methods were thus able to clearly distinguish the flow patterns of CCA and CB in systole. Vector concentration from angle-independent vector velocity estimates is a quantitative index, which is simple to calculate and can differentiate between laminar and complex flow. (E-mail: [email protected]) Ó 2014 World Federation for Ultrasound in Medicine & Biology. Key Words: Ultrasound, Vector flow, Transverse oscillation, Carotid artery.
velocity (Caro et al. 1969; Fuster et al. 1990; Glagov et al. 1988; Heijenbrok-Kal et al. 2006; Hugh and Fox 1970; Peterson et al. 1960; Saba 2010; Saba 2012; Zarins et al. 1983). Others have investigated the turbulence of blood flow in diseased vessels by estimation of the peak Doppler frequency and spectral broadening index (Brown et al. 1982; Kassam et al. 1982) and by Doppler velocity analysis (Phillips et al. 1983). However, spectral Doppler only estimates the axial velocity component, and therefore the complexity of blood flow remains hidden in conventional US examinations (Evans et al. 2011; Jensen 1996; Jensen and Munk 1998). With the vector US technique transverse oscillation (TO) the blood velocities of both the axial and the transverse directions are obtained and the complexity of blood flow can be visualized and quantified (Jensen
INTRODUCTION Carotid artery stenosis as a result of atherothrombotic disease may lead to ischemic stroke, which is one of the leading causes of mortality and morbidity in western countries (Murray and Lopez 1997; Sacco and Adams 2006). Several modalities are used to characterize the luminal narrowing and the morphology of the atherosclerotic plaques, such as ultrasound (US), computed tomography (CT) and magnetic resonance imaging (Saba 2012). Dynamic changes in the carotid artery are mainly assessed by spectral Doppler estimation of the peak systolic
Address correspondence to: Dr. Mads Møller Pedersen, Department of Radiology, University of Copenhagen, Blegdamsvej 9, 2100 Copenhagen, Denmark. E-mail: [email protected] 2700
Vector flow quantification of the carotid artery d M. M. PEDERSEN et al.
2701
Fig. 1. Image frame demonstrating vector flow ultrasound at the carotid bifurcation using a commercial ultrasound scanner. The common carotid artery is to the right, the carotid bulb to the left. The white vectors illustrate the velocity, magnitude and direction of the blood flow. The color scale used is inserted in the upper right corner for illustration. The vectors were used for the visual assessment method, the colored pixels for the quantification method.
2001; Jensen and Munk 1998; Munk 1996). The technique introduces an ultrasound field oscillating in both the axial and lateral direction. This can be used for estimating both the axial and lateral velocity components independently. Changes in severity of stenosis produced by altered pressure and flow may be important in the pathogenesis of carotid artery stenosis. In this study we propose a new quantification method, vector concentration, a unique method for imaging carotid abnormalities and dynamic changes. The aim of the study was to determine the technical performance and interpretation of vector concentration as a tool for estimation of flow complexity. A secondary aim was to establish accuracy parameters to detect flow changes/patterns in the common carotid artery (CCA) and the carotid bulb (CB).
MATERIALS AND METHODS Patients This prospective study was performed after approval by the National Committee on Biomedical Research Ethics. Eight healthy volunteers (5 men, 3 women; mean age 39.5 y, range 28–45 y) were included in the study after verbal informed consent as approved by the ethics commettee. The verbal consents performed by a medical doctor were not documented.
The transverse oscillation method The basis of the TO method has previously been described (Jensen 2001; Jensen and Munk 1998; Munk 1996). In brief, the TO method tracks scatterer motion along two orthogonal axes by emitting a pulse identical to conventional Doppler US. The motion in the axial and the transverse directions are found by generating a transverse oscillating field and the use of two special estimators for finding the axial and lateral velocity components (Jensen 2001). The 2-D vector velocity of the moving scatterer is found by combining the velocity components along the two axes. TO has been validated in a flow-rig (Udesen and Jensen 2006), against magnetic resonance angiography (Hansen et al. 2009a, 2009b) and conventional spectral Doppler US (Pedersen et al. 2012). Data acquisition The right carotid bifurcation, including the CCA and CB, of all eight volunteers was scanned in a longitudinal plane with vector flow US as shown in Figure 1. All scan sequences were acquired with a commercial US scanner (ProFocus Ultraview, BK Medical, Herlev, Denmark) and a 4–12 MHz linear transducer (8670, BK Medical). The TO data were visualized in real time with colorencoded vector data for each pixel, where a circular color map is used to decode the velocity and direction of the
2702
Ultrasound in Medicine and Biology
Volume 40, Number 11, 2014
Fig. 2. Demonstration of an image frame used for visual assessment by medical experts. The cropped image included two outlined areas. The green boxes define the common carotid area (right) and carotid bulb (left). Each of the 164 frames available from the eight volunteers was presented randomly to the experts using a web interface.
flow. White vectors can also be enabled to indicate the direction and velocity magnitude. Using a frame rate of 22 frames per second, one cardiac cycle was obtained covering 17, 19, 22, 22, 20, 19, 22 and 23 image frames for each individual volunteer, respectively. All the frames for all the selected cardiac cycles were used for further evaluation. The number of frames because of different heart rates. A total of 164 image frames from both diastole and systole were used in this study, where the CCA and CB areas were outlined on every image (Fig. 2). These areas were used for the evaluation by the medical experts and for quantitative vector calculation. Expert evaluation An in-house personal home page web interface and a My Structured Query Language open-source rational database management system were used to present the 164 images in random order and to store the evaluations. Each frame was presented to five medical experts (radiologists) who evaluated whether complex flow was present (score 0) or not (score 1) in each of the two areas (Fig. 2). Complex flow was defined as a more than 90 degree change in vector angles within the area. A simple all or nothing decision was chosen to make sure all experts evaluated from the same definition. For each volunteer the mean expert evaluation value and the standard deviation were calculated for CCA and CB.
Vector data decoding The axial and transverse vector velocity magnitudes of the image frames encoded as colored pixels (Fig. 1) were decoded into the axial velocity magnitudes, vz, and the transverse velocity magnitudes, vx, using an inhouse Matlab (Mathworks, Natick, MA, USA) script. A 2-dimensional (2-D) vector field was calculated for each of the two areas in each image frame. Vector concentration of a 2-D vector field Vector concentration is a single metric describing the complex flow and is previously described by Batschelet (1981). An area with complex flow contains subareas with different flow directions and will thus not have a dominant direction. It can be compared to a freeway with cars driving in different directions. The vector concentration is an expression of the spread in direction. In the opposite situation an area with subareas of uniform directions will have a low spread and thus a high vector concentration. The method provides a number between 0 and 1, where 1 represents no complexity and 0 fully complex flow. The vector concentration is found as described here: For a vector at position i, the flow angle, qi, was calculated as qi 5 arctanðvx;i ; vz;i Þ;
(1)
Each vector angle, qi, is represented on the unit circle as pi 5 (xi, yi), where xi 5 cos (qi) and yi 5 sin (qi)
Vector flow quantification of the carotid artery d M. M. PEDERSEN et al.
2703
Table 2. Quantification of flow complexity in the right common carotid artery and carotid bulb in one cardiac cycle of eight healthy volunteers. The mean vector concentrations and standard deviations were calculated Mean r 6 std
Fig. 3. Three unit circle plots. (a) Example, where the vector concentration, r (red line), is calculated for a vector field of two vectors, p1 (green) and p2 (blue). The mean x and y values are used to calculate the distance to r, which is the vector concentration. (b) Full complex flow, where r 5 0 for a vector field of five vectors. (c) The vector concentration increases when all points are concentrated in one part of the circle.
(Fig. 3). The mean values of all points were calculated for each 2-D vector field as n 1 X x5 , cosðqi Þ; n i51
(2)
n 1 X y5 , sinðqi Þ; n i51
(3)
Table 1. Visual assessment of flow complexity in the right common carotid artery and carotid bulb in one cardiac cycle of eight healthy volunteers. Five ultrasound experts individually scored all 164 frames separately in random order for the presence (score 0), or absence (score 1) of complex flow Mean visual evaluation 6 1 standard deviation Volunteer
Common carotid artery
Carotid bulb
1 2 3 4 5 6 7 8 Mean 6 std
0.87 6 0.33 0.89 6 0.31 0.57 6 0.50 0.95 6 0.23 0.83 6 0.38 0.62 6 0.49 0.87 6 0.33 0.95 6 0.22 0.82 6 0.14
0.21 6 0.41 0.35 6 0.48 0.05 6 0.21 0.39 6 0.49 0.10 6 0.30 0.06 6 0.24 0.06 6 0.25 0.65 6 0.48 0.23 6 0.22
Volunteer
Common carotid artery
Carotid bulb
1 2 3 4 5 6 7 8 Mean 6 std
0.97 6 0.22 0.96 6 0.24 0.94 6 0.34 0.97 6 0.22 0.96 6 0.22 0.94 6 0.35 0.98 6 0.17 0.98 6 0.17 0.96 6 0.16
0.90 6 0.39 0.88 6 0.46 0.67 6 0.78 0.80 6 0.61 0.84 6 0.56 0.83 6 0.57 0.79 6 0.63 0.89 6 0.43 0.83 6 0.07
where n is the total number of vectors in the 2-D vector field. Simple circular statistics was used to determine the flow complexity. The resulting vector concentration, r, was found using Pythagoras’ theorem, qffiffiffiffiffiffiffiffiffiffiffiffi (4) r 5 x2 1y2 ; as illustrated in Figure 3. Turbulent or disturbed flow is present when there is some obstruction that results in a disruption of the normal laminar pattern. For a perfectly laminar blood flow patterns, r equals 1, whereas r decreases toward a value of zero with increasing complexity as described by Batschelet (1981). Statistical analysis For each image frame the sum of all CCA and CB evaluations were used in a Fleiss k analysis to calculate the level of inter-observer agreement, k, where k 5 [0.41 0.60] was interpreted as moderate agreement. The mean expert evaluations and the mean vector concentrations for CCA and CB were tested with a Wilcoxon’s matched pair test, and p , 0.05 was considered significant. All statistical analyses were performed using Matlab. RESULTS Vector flow US imaging of CCA and CB were evaluated visually and quantitatively. The results of the visual evaluation by medical experts and the flow quantification using vector concentration are shown in Tables 1 and 2, respectively. The medical experts evaluated the flow to be mainly laminar in CCA (0.82 6 0.14) and mainly complex (0.23 6 0.22) in CB. Likewise, the estimated vector concentrations in CCA (0.96 6 0.16) indicated mainly laminar flow and in CB (0.83 6 0.07) indicated mainly turbulence. Blood flow dynamics of one single cardiac cycle and its relationship to calculated vector concentrations in one
2704
Ultrasound in Medicine and Biology
Volume 40, Number 11, 2014
Fig. 4. Vector flow data of the carotid bifurcation for volunteer 2. The vector concentration as a function of time (top), and two image frames of different (bottom left), and similar (bottom right) r values for the common carotid artery (green), and the carotid bulb (red).
volunteer is shown in Figure 4. An overview for all volunteers is shown in Figure 5. At peak systole the vector concentration values clearly differentiated the CCA from CB. The Fleiss k test indicated moderate inter-observer agreement between the visual evaluations for CCA (k 5 0.42) and for CB (k 5 0.52), which is indicated in Table 1 from the calculated standard deviations. The Wilcoxon’s matched pair test rejected the null hypothesis for visual evaluation (p 5 0.0078) and mean vector concentration (p 5 0.0078), indicating that both approaches found significantly different flow patterns in CCA compared with CB for all the cases studied. A linear regression analysis of the visual evaluations as a function of the vector concentration for CCA and CB (Fig. 6) resulted in an r value of 0.86, suggesting a linear correlation. DISCUSSION In this study of healthy volunteers a vortex with complex flow was found in carotid bulbs, whereas the CCA had mainly laminar flow. For both methods, visual evaluation by medical experts and the automated vector concentration method are able to clearly distinguish the flow patterns of CCA and CB in systole. Vector concentration from angle-independent vector velocity estimates is a quantitative index that is simple to calculate and automate. Complex flow in healthy volunteers has previously
been investigated with US vector flow techniques in different vascular sites (Hansen 2009a; Hansen et al. 2011; Udesen et al. 2007). However, so far no attempts have been made to measure the flow complexity, and this is, to our knowledge, the first in vivo study, where the flow patterns in the CCA and CB are evaluated quantitatively with the vector flow US technique. Complex flow assessed visually had only a moderate agreement when studied by different US experts. Studies of regurgitation and turbulence evaluated with conventional color flow mapping have reported inter-observer and intra-observer variability of 20% agreements (Castello et al. 1992; Hoit et al. 1989). On the other hand, the quantitative vector concentration is an automated approach using angle-independent vector velocity estimates and requires only an operator determination of the anatomic area of interest. The linear regression analysis with an r value of 0.86 indicates that there is a linear relation between the visual evaluation and the vector concentration. This value is suggested to be calculated in future clinical studies of CCA and CB in atherosclerotic patients. The visual evaluations were performed using vector velocity maps, which is a limitation in our study. The experts are not familiar with this kind of presentation of flow patterns and thus it is a confounder when assessing the variability between observers. Second, colored vector information is not totally depicted during the entire cardiac cycle, which is another drawback in our study. This is
Vector flow quantification of the carotid artery d M. M. PEDERSEN et al.
2705
Fig. 5. The vector concentration throughout one cardiac cycle for all volunteers. The values of the common carotid artery (green) and the carotid bulb (red), and the threshold value are shown.
Fig. 6. The mean expert evaluation as a function of the mean vector concentration for the common carotid artery and the carotid bulb of all volunteers. A linear regression line was calculated with an r value of 0.86.
primarily affected by the pulse repetition frequency, the echo-canceling filter and the wall filter. To avoid aliasing, the settings had to be optimized to allow correct measurements at peak velocities during systole and thus measurements of slower flow during the diastole were compromised. This is illustrated in Figure 4, where frames are shown at systole (left frame) and diastole (right frame). For the expert evaluation of the flow patterns, a simple all or nothing score was used. A continuous evaluation scale from 0 to 1 could have been used. However, this would require that the experts should use a more complex definition of the flow patterns on such scale. Further analysis of the inter-observer differences was not performed. In future studies the variation caused by inter-observer differences should be approached further. The anatomic areas used for defining the vessels varied in size, which for the visual evaluations may have an impact on the judgment in larger areas compared with the smaller areas. Boxes of equal areas would therefore have balanced the visual evaluations. However, the acquired
2706
Ultrasound in Medicine and Biology
images containing both the CCA and the CB did not allow for boxes of equal area because the imaged part of the CCA differed in size compared with the CB. The differences were accounted for in the vector concentration calculation by dividing with the number of measurements (Figs. 2 and 3). The vector concentration index has higher values for complex flow than for the human observers. This is due to the difference in calculating the indices. The vector concentration is an average over an area, whereas the human index is a single number. For a completely random flow, the vector concentration can reach zero, but most often this is only the case in a short part of the cardiac cycle. It was assumed that the flow patterns in the CCA and the CB differ at all times during the cardiac cycle. However, the flow patterns of the two areas appear most obvious at peak systole. Thus, comparison between selected frames at peak systole should be considered in future studies. However, the ultrasound scanner was not equipped with cardiac gating hardware and thus it was not possible to define the cardiac cycle further. The flow patterns presented to the experts were TO images. A further visual evaluation using standard color flow Doppler imaging should be considered in future studies. It is technically possible to create conventional flow Doppler images from the TO data set using only the axial velocities. It is well known that the incidence of cardiovascular events correlates with surrogate markers for atherosclerosis such as carotid artery intima media thickness. Similar, the vector concentration assessment might be of clinical interest in regard to subclinical vascular damage. The present observations suggest that the vector flow technique represents, at least in part, a tool for early detection of adverse vascular wall processes. In conclusion, this study has for the first time compared visual evaluations with a quantitative method of vector flow ultrasound of the CCA and CB. A linear relation is suggested and during systole the vector concentration clearly distinguishes the flow patterns of CCA and CB. Both visual evaluation and the automated vector concentration method are able to distinguish the flow patterns in the CCA and CB. REFERENCES Batschelet E. Circular statistics in biology. London: Academic Press; 1981. Brown PM, Johnston KW, Kassam M, Cobbold RS. A critical study of ultrasound Doppler spectral analysis for detecting carotid disease. Ultrasound Med Biol 1982;5:515–523. Caro CG, Fitz-Gerald JM, Schroter RC. Arterial wall shear and distribution of early atheroma in man. Nature 1969;223:1159–1160. Castello R, Lenzen P, Aguirre F, Labovitz A. Variability in the quantitation of mitral regurgitation by Doppler color flow mapping: Comparison of transthoracic and transesophageal studies. J Am Coll Cardiol 1992;20:433–438.
Volume 40, Number 11, 2014 Evans DH, Jensen JA, Nielsen MB. Ultrasonic colour Doppler imaging. Interface Focus 2011;1:490–502. Fuster V, Stein B, Ambrose JA, Badimon L, Badimon JJ, Chesebro JH. Atherosclerotic plaque rupture and thrombosis. Evolving concepts. Circulation 1990;82(3 Suppl):II47–II59. Glagov S, Zarins CK, Giddens DP, Ku DN. Hemodynamics and atherosclerosis. Insights and perspectives gained from studies of human arteries. Arch Pathol Lab Med 1988;112:1018–1031. Hansen KL, Udesen J, Gran F, Jensen JA, Bachmann Nielsen M. In-vivo examples of flow patterns with the fast vector velocity ultrasound method. Ultraschall Med 2009a;30:471–477. Hansen KL, Udesen J, Oddershede N, Henze L, Thomsen C, Jensen JA, Nielsen MB. In vivo comparison of three ultrasound vector velocity techniques to MR phase contrast angiography. Ultrasonics 2009b; 49:659–667. Hansen KL, Udesen J, Thomsen C, Jensen JA, Nielsen MB. In vivo validation of a blood vector velocity estimator with MR angiography. IEEE Trans Ultrason Ferroelec Freq Contr 2009c;56:91–100. Hansen PM, Pedersen MM, Hansen KL, Nielsen MB, Jensen JA. New technology—demonstration of a vector velocity technique. Ultraschall Med 2011;32:213–215. Heijenbrok-Kal MH, Buskens E, Naderkoorn PJ, van der Graaf Y, Hunink MGM. Optimal peak systolic velocity threshold at duplex US for determining the need for carotid endarterectomy: A decision analytic approach. Radiology 2006;238:480–488. Hoit BD, Jones M, Eidbo EE, Elias W, Sahn DJ. Sources of variability for Doppler color flow mapping of regurgitant jets in an animal model of mitral regurgitation. J Am Coll Cardiol 1989;13: 1631–1636. Hugh AE, Fox JA. The precise localisation of atheroma and its association with stasis at the origin of the internal carotid artery—a radiographic investigation. Br J Radiol 1970;43:377–383. Jensen JA. Estimation of blood velocities using ultrasound: A signal processing approach. New York: Cambridge University Press; 1996. Jensen JA. A new estimator for vector velocity estimation. IEEE Trans Ultrason Ferroelec Freq Contr 2001;48:886–894. Jensen JA, Munk P. A new method for estimation of velocity vectors. IEEE Trans Ultrason Ferroelec Freq Contr 1998;45:837–851. Kassam MS, Cobbold RS, Johnston KW, Graham CM. Method for estimating the Doppler mean velocity waveform. Ultrasound Med Biol 1982;8:537–544. Munk P. Estimation of the 2-D flow vector in ultrasonic imaging: A new approach. Master’s thesis. Copenhagen: Technical University of Denmark, 1996. Murray CJ, Lopez AD. Alternative projections of mortality and disability by cause 1990-2020: Global burden of disease study. Lancet 1997;349:1498–1504. Pedersen MM, Pihl MJ, Haugaard P, Hansen JM, Hansen KL, Nielsen MB, Jensen JA. Comparison of real-time in vivo spectral and vector velocity estimation. Ultrasound Med Biol 2012;38:145–151. Peterson RE, Livingston KE, Escobar A. Development and distribution of gross atherosclerotic lesions at cervical carotid bifurcation. Neurology 1960;10:955–959. Phillips DJ, Greene FM, Langlois Y, Roederer GO, Strandness DE. Flow velocity patterns in the carotid bifurcations of young, presumed normal subjects. Ultrasound Med Biol 1983;9:39–49. Saba L, Sanfilippo R, Montisci R, Mallarini G. Correlation between USPSVand MDCTA in the quantification of carotid artery stenosis. Eur J Radiol 2010;74:99–103. Saba L. Imaging of the carotid artery. Atherosclerosis 2012;220:294–309. Sacco RL, Adams R. Albers. Guidelines for prevention of stroke in patients with ischemic stroke or transient ischemic attack. Stroke 2006; 37:577–617. Udesen J, Jensen JA. Investigation of transverse oscillation method. IEEE Trans Ultrason Ferroelec Freq Contr 2006;53:959–971. Udesen J, Nielsen MB, Nielsen KR, Jensen JA. Examples of in vivoblood vector velocity estimation. Ultrasound Med Biol 2007;33:541–548. Zarins CK, Giddens DP, Bharadvaj BK, Sottiurai VS, Mabon RF, Glagov S. Carotid bifurcation atherosclerosis. Quantitative correlation of plaque localization with flow velocity profiles and wall shear stress. Circ Res 1983;53:502–514.
http://dx.doi.org/10.1016/j.ultrasmedbio.2014.06.001
d
Original Contribution NOVEL FLOW QUANTIFICATION OF THE CAROTID BULB AND THE COMMON CAROTID ARTERY WITH VECTOR FLOW ULTRASOUND MADS MØLLER PEDERSEN,* MICHAEL JOHANNES PIHL,y PER HAUGAARD,z k KRISTOFFER LINDSKOV HANSEN,* THEIS LANGE,x LARS L€ ONN,* MICHAEL BACHMANN NIELSEN,* and JØRGEN ARENDT JENSENy * Department of Radiology, Rigshospitalet, University of Copenhagen, Copenhagen, Denmark; y Center for Fast Ultrasound Imaging, Department of Electrical Engeneering, Technical University of Denmark, Lyngby, Denmark; z B-K Medical, Herlev, Denmark; x Department of Biostatistics, University of Copenhagen, Copenhagen, Denmark; and k Department of Vascular Surgery, Rigshospitalet, University of Copenhagen, Copenhagen, Denmark (Received 6 December 2013; revised 2 April 2014; in final form 3 June 2014)
Abstract—Abnormal blood flow is usually assessed using spectral Doppler estimation of the peak systolic velocity. The technique, however, only estimates the axial velocity component, and therefore the complexity of blood flow remains hidden in conventional ultrasound examinations. With the vector ultrasound technique transverse oscillation the blood velocities of both the axial and the transverse directions are obtained and the complexity of blood flow can be visualized. The aim of the study was to determine the technical performance and interpretation of vector concentration as a tool for estimation of flow complexity. A secondary aim was to establish accuracy parameters to detect flow changes/patterns in the common carotid artery (CCA) and the carotid bulb (CB). The right carotid bifurcation including the CCA and CB of eight healthy volunteers were scanned in a longitudinal plane with vector flow ultrasound (US) using a commercial vector flow ultrasound scanner (ProFocus, BK Medical, Denmark) with a linear 5 MHz transducer transverse oscillation vector flow software. CCA and CB areas were marked in one cardiac cycle from each volunteer. The complex flow was assessed by medical expert evaluation and by vector concentration calculation. A vortex with complex flow was found in all carotid bulbs, whereas the CCA had mainly laminar flow. The medical experts evaluated the flow to be mainly laminar in the CCA (0.82 ± 0.14) and mainly complex (0.23 ± 0.22) in the CB. Likewise, the estimated vector concentrations in CCA (0.96 ± 0.16) indicated mainly laminar flow and in CB (0.83 ± 0.07) indicated mainly turbulence. Both methods were thus able to clearly distinguish the flow patterns of CCA and CB in systole. Vector concentration from angle-independent vector velocity estimates is a quantitative index, which is simple to calculate and can differentiate between laminar and complex flow. (E-mail: [email protected]) Ó 2014 World Federation for Ultrasound in Medicine & Biology. Key Words: Ultrasound, Vector flow, Transverse oscillation, Carotid artery.
velocity (Caro et al. 1969; Fuster et al. 1990; Glagov et al. 1988; Heijenbrok-Kal et al. 2006; Hugh and Fox 1970; Peterson et al. 1960; Saba 2010; Saba 2012; Zarins et al. 1983). Others have investigated the turbulence of blood flow in diseased vessels by estimation of the peak Doppler frequency and spectral broadening index (Brown et al. 1982; Kassam et al. 1982) and by Doppler velocity analysis (Phillips et al. 1983). However, spectral Doppler only estimates the axial velocity component, and therefore the complexity of blood flow remains hidden in conventional US examinations (Evans et al. 2011; Jensen 1996; Jensen and Munk 1998). With the vector US technique transverse oscillation (TO) the blood velocities of both the axial and the transverse directions are obtained and the complexity of blood flow can be visualized and quantified (Jensen
INTRODUCTION Carotid artery stenosis as a result of atherothrombotic disease may lead to ischemic stroke, which is one of the leading causes of mortality and morbidity in western countries (Murray and Lopez 1997; Sacco and Adams 2006). Several modalities are used to characterize the luminal narrowing and the morphology of the atherosclerotic plaques, such as ultrasound (US), computed tomography (CT) and magnetic resonance imaging (Saba 2012). Dynamic changes in the carotid artery are mainly assessed by spectral Doppler estimation of the peak systolic
Address correspondence to: Dr. Mads Møller Pedersen, Department of Radiology, University of Copenhagen, Blegdamsvej 9, 2100 Copenhagen, Denmark. E-mail: [email protected] 2700
Vector flow quantification of the carotid artery d M. M. PEDERSEN et al.
2701
Fig. 1. Image frame demonstrating vector flow ultrasound at the carotid bifurcation using a commercial ultrasound scanner. The common carotid artery is to the right, the carotid bulb to the left. The white vectors illustrate the velocity, magnitude and direction of the blood flow. The color scale used is inserted in the upper right corner for illustration. The vectors were used for the visual assessment method, the colored pixels for the quantification method.
2001; Jensen and Munk 1998; Munk 1996). The technique introduces an ultrasound field oscillating in both the axial and lateral direction. This can be used for estimating both the axial and lateral velocity components independently. Changes in severity of stenosis produced by altered pressure and flow may be important in the pathogenesis of carotid artery stenosis. In this study we propose a new quantification method, vector concentration, a unique method for imaging carotid abnormalities and dynamic changes. The aim of the study was to determine the technical performance and interpretation of vector concentration as a tool for estimation of flow complexity. A secondary aim was to establish accuracy parameters to detect flow changes/patterns in the common carotid artery (CCA) and the carotid bulb (CB).
MATERIALS AND METHODS Patients This prospective study was performed after approval by the National Committee on Biomedical Research Ethics. Eight healthy volunteers (5 men, 3 women; mean age 39.5 y, range 28–45 y) were included in the study after verbal informed consent as approved by the ethics commettee. The verbal consents performed by a medical doctor were not documented.
The transverse oscillation method The basis of the TO method has previously been described (Jensen 2001; Jensen and Munk 1998; Munk 1996). In brief, the TO method tracks scatterer motion along two orthogonal axes by emitting a pulse identical to conventional Doppler US. The motion in the axial and the transverse directions are found by generating a transverse oscillating field and the use of two special estimators for finding the axial and lateral velocity components (Jensen 2001). The 2-D vector velocity of the moving scatterer is found by combining the velocity components along the two axes. TO has been validated in a flow-rig (Udesen and Jensen 2006), against magnetic resonance angiography (Hansen et al. 2009a, 2009b) and conventional spectral Doppler US (Pedersen et al. 2012). Data acquisition The right carotid bifurcation, including the CCA and CB, of all eight volunteers was scanned in a longitudinal plane with vector flow US as shown in Figure 1. All scan sequences were acquired with a commercial US scanner (ProFocus Ultraview, BK Medical, Herlev, Denmark) and a 4–12 MHz linear transducer (8670, BK Medical). The TO data were visualized in real time with colorencoded vector data for each pixel, where a circular color map is used to decode the velocity and direction of the
2702
Ultrasound in Medicine and Biology
Volume 40, Number 11, 2014
Fig. 2. Demonstration of an image frame used for visual assessment by medical experts. The cropped image included two outlined areas. The green boxes define the common carotid area (right) and carotid bulb (left). Each of the 164 frames available from the eight volunteers was presented randomly to the experts using a web interface.
flow. White vectors can also be enabled to indicate the direction and velocity magnitude. Using a frame rate of 22 frames per second, one cardiac cycle was obtained covering 17, 19, 22, 22, 20, 19, 22 and 23 image frames for each individual volunteer, respectively. All the frames for all the selected cardiac cycles were used for further evaluation. The number of frames because of different heart rates. A total of 164 image frames from both diastole and systole were used in this study, where the CCA and CB areas were outlined on every image (Fig. 2). These areas were used for the evaluation by the medical experts and for quantitative vector calculation. Expert evaluation An in-house personal home page web interface and a My Structured Query Language open-source rational database management system were used to present the 164 images in random order and to store the evaluations. Each frame was presented to five medical experts (radiologists) who evaluated whether complex flow was present (score 0) or not (score 1) in each of the two areas (Fig. 2). Complex flow was defined as a more than 90 degree change in vector angles within the area. A simple all or nothing decision was chosen to make sure all experts evaluated from the same definition. For each volunteer the mean expert evaluation value and the standard deviation were calculated for CCA and CB.
Vector data decoding The axial and transverse vector velocity magnitudes of the image frames encoded as colored pixels (Fig. 1) were decoded into the axial velocity magnitudes, vz, and the transverse velocity magnitudes, vx, using an inhouse Matlab (Mathworks, Natick, MA, USA) script. A 2-dimensional (2-D) vector field was calculated for each of the two areas in each image frame. Vector concentration of a 2-D vector field Vector concentration is a single metric describing the complex flow and is previously described by Batschelet (1981). An area with complex flow contains subareas with different flow directions and will thus not have a dominant direction. It can be compared to a freeway with cars driving in different directions. The vector concentration is an expression of the spread in direction. In the opposite situation an area with subareas of uniform directions will have a low spread and thus a high vector concentration. The method provides a number between 0 and 1, where 1 represents no complexity and 0 fully complex flow. The vector concentration is found as described here: For a vector at position i, the flow angle, qi, was calculated as qi 5 arctanðvx;i ; vz;i Þ;
(1)
Each vector angle, qi, is represented on the unit circle as pi 5 (xi, yi), where xi 5 cos (qi) and yi 5 sin (qi)
Vector flow quantification of the carotid artery d M. M. PEDERSEN et al.
2703
Table 2. Quantification of flow complexity in the right common carotid artery and carotid bulb in one cardiac cycle of eight healthy volunteers. The mean vector concentrations and standard deviations were calculated Mean r 6 std
Fig. 3. Three unit circle plots. (a) Example, where the vector concentration, r (red line), is calculated for a vector field of two vectors, p1 (green) and p2 (blue). The mean x and y values are used to calculate the distance to r, which is the vector concentration. (b) Full complex flow, where r 5 0 for a vector field of five vectors. (c) The vector concentration increases when all points are concentrated in one part of the circle.
(Fig. 3). The mean values of all points were calculated for each 2-D vector field as n 1 X x5 , cosðqi Þ; n i51
(2)
n 1 X y5 , sinðqi Þ; n i51
(3)
Table 1. Visual assessment of flow complexity in the right common carotid artery and carotid bulb in one cardiac cycle of eight healthy volunteers. Five ultrasound experts individually scored all 164 frames separately in random order for the presence (score 0), or absence (score 1) of complex flow Mean visual evaluation 6 1 standard deviation Volunteer
Common carotid artery
Carotid bulb
1 2 3 4 5 6 7 8 Mean 6 std
0.87 6 0.33 0.89 6 0.31 0.57 6 0.50 0.95 6 0.23 0.83 6 0.38 0.62 6 0.49 0.87 6 0.33 0.95 6 0.22 0.82 6 0.14
0.21 6 0.41 0.35 6 0.48 0.05 6 0.21 0.39 6 0.49 0.10 6 0.30 0.06 6 0.24 0.06 6 0.25 0.65 6 0.48 0.23 6 0.22
Volunteer
Common carotid artery
Carotid bulb
1 2 3 4 5 6 7 8 Mean 6 std
0.97 6 0.22 0.96 6 0.24 0.94 6 0.34 0.97 6 0.22 0.96 6 0.22 0.94 6 0.35 0.98 6 0.17 0.98 6 0.17 0.96 6 0.16
0.90 6 0.39 0.88 6 0.46 0.67 6 0.78 0.80 6 0.61 0.84 6 0.56 0.83 6 0.57 0.79 6 0.63 0.89 6 0.43 0.83 6 0.07
where n is the total number of vectors in the 2-D vector field. Simple circular statistics was used to determine the flow complexity. The resulting vector concentration, r, was found using Pythagoras’ theorem, qffiffiffiffiffiffiffiffiffiffiffiffi (4) r 5 x2 1y2 ; as illustrated in Figure 3. Turbulent or disturbed flow is present when there is some obstruction that results in a disruption of the normal laminar pattern. For a perfectly laminar blood flow patterns, r equals 1, whereas r decreases toward a value of zero with increasing complexity as described by Batschelet (1981). Statistical analysis For each image frame the sum of all CCA and CB evaluations were used in a Fleiss k analysis to calculate the level of inter-observer agreement, k, where k 5 [0.41 0.60] was interpreted as moderate agreement. The mean expert evaluations and the mean vector concentrations for CCA and CB were tested with a Wilcoxon’s matched pair test, and p , 0.05 was considered significant. All statistical analyses were performed using Matlab. RESULTS Vector flow US imaging of CCA and CB were evaluated visually and quantitatively. The results of the visual evaluation by medical experts and the flow quantification using vector concentration are shown in Tables 1 and 2, respectively. The medical experts evaluated the flow to be mainly laminar in CCA (0.82 6 0.14) and mainly complex (0.23 6 0.22) in CB. Likewise, the estimated vector concentrations in CCA (0.96 6 0.16) indicated mainly laminar flow and in CB (0.83 6 0.07) indicated mainly turbulence. Blood flow dynamics of one single cardiac cycle and its relationship to calculated vector concentrations in one
2704
Ultrasound in Medicine and Biology
Volume 40, Number 11, 2014
Fig. 4. Vector flow data of the carotid bifurcation for volunteer 2. The vector concentration as a function of time (top), and two image frames of different (bottom left), and similar (bottom right) r values for the common carotid artery (green), and the carotid bulb (red).
volunteer is shown in Figure 4. An overview for all volunteers is shown in Figure 5. At peak systole the vector concentration values clearly differentiated the CCA from CB. The Fleiss k test indicated moderate inter-observer agreement between the visual evaluations for CCA (k 5 0.42) and for CB (k 5 0.52), which is indicated in Table 1 from the calculated standard deviations. The Wilcoxon’s matched pair test rejected the null hypothesis for visual evaluation (p 5 0.0078) and mean vector concentration (p 5 0.0078), indicating that both approaches found significantly different flow patterns in CCA compared with CB for all the cases studied. A linear regression analysis of the visual evaluations as a function of the vector concentration for CCA and CB (Fig. 6) resulted in an r value of 0.86, suggesting a linear correlation. DISCUSSION In this study of healthy volunteers a vortex with complex flow was found in carotid bulbs, whereas the CCA had mainly laminar flow. For both methods, visual evaluation by medical experts and the automated vector concentration method are able to clearly distinguish the flow patterns of CCA and CB in systole. Vector concentration from angle-independent vector velocity estimates is a quantitative index that is simple to calculate and automate. Complex flow in healthy volunteers has previously
been investigated with US vector flow techniques in different vascular sites (Hansen 2009a; Hansen et al. 2011; Udesen et al. 2007). However, so far no attempts have been made to measure the flow complexity, and this is, to our knowledge, the first in vivo study, where the flow patterns in the CCA and CB are evaluated quantitatively with the vector flow US technique. Complex flow assessed visually had only a moderate agreement when studied by different US experts. Studies of regurgitation and turbulence evaluated with conventional color flow mapping have reported inter-observer and intra-observer variability of 20% agreements (Castello et al. 1992; Hoit et al. 1989). On the other hand, the quantitative vector concentration is an automated approach using angle-independent vector velocity estimates and requires only an operator determination of the anatomic area of interest. The linear regression analysis with an r value of 0.86 indicates that there is a linear relation between the visual evaluation and the vector concentration. This value is suggested to be calculated in future clinical studies of CCA and CB in atherosclerotic patients. The visual evaluations were performed using vector velocity maps, which is a limitation in our study. The experts are not familiar with this kind of presentation of flow patterns and thus it is a confounder when assessing the variability between observers. Second, colored vector information is not totally depicted during the entire cardiac cycle, which is another drawback in our study. This is
Vector flow quantification of the carotid artery d M. M. PEDERSEN et al.
2705
Fig. 5. The vector concentration throughout one cardiac cycle for all volunteers. The values of the common carotid artery (green) and the carotid bulb (red), and the threshold value are shown.
Fig. 6. The mean expert evaluation as a function of the mean vector concentration for the common carotid artery and the carotid bulb of all volunteers. A linear regression line was calculated with an r value of 0.86.
primarily affected by the pulse repetition frequency, the echo-canceling filter and the wall filter. To avoid aliasing, the settings had to be optimized to allow correct measurements at peak velocities during systole and thus measurements of slower flow during the diastole were compromised. This is illustrated in Figure 4, where frames are shown at systole (left frame) and diastole (right frame). For the expert evaluation of the flow patterns, a simple all or nothing score was used. A continuous evaluation scale from 0 to 1 could have been used. However, this would require that the experts should use a more complex definition of the flow patterns on such scale. Further analysis of the inter-observer differences was not performed. In future studies the variation caused by inter-observer differences should be approached further. The anatomic areas used for defining the vessels varied in size, which for the visual evaluations may have an impact on the judgment in larger areas compared with the smaller areas. Boxes of equal areas would therefore have balanced the visual evaluations. However, the acquired
2706
Ultrasound in Medicine and Biology
images containing both the CCA and the CB did not allow for boxes of equal area because the imaged part of the CCA differed in size compared with the CB. The differences were accounted for in the vector concentration calculation by dividing with the number of measurements (Figs. 2 and 3). The vector concentration index has higher values for complex flow than for the human observers. This is due to the difference in calculating the indices. The vector concentration is an average over an area, whereas the human index is a single number. For a completely random flow, the vector concentration can reach zero, but most often this is only the case in a short part of the cardiac cycle. It was assumed that the flow patterns in the CCA and the CB differ at all times during the cardiac cycle. However, the flow patterns of the two areas appear most obvious at peak systole. Thus, comparison between selected frames at peak systole should be considered in future studies. However, the ultrasound scanner was not equipped with cardiac gating hardware and thus it was not possible to define the cardiac cycle further. The flow patterns presented to the experts were TO images. A further visual evaluation using standard color flow Doppler imaging should be considered in future studies. It is technically possible to create conventional flow Doppler images from the TO data set using only the axial velocities. It is well known that the incidence of cardiovascular events correlates with surrogate markers for atherosclerosis such as carotid artery intima media thickness. Similar, the vector concentration assessment might be of clinical interest in regard to subclinical vascular damage. The present observations suggest that the vector flow technique represents, at least in part, a tool for early detection of adverse vascular wall processes. In conclusion, this study has for the first time compared visual evaluations with a quantitative method of vector flow ultrasound of the CCA and CB. A linear relation is suggested and during systole the vector concentration clearly distinguishes the flow patterns of CCA and CB. Both visual evaluation and the automated vector concentration method are able to distinguish the flow patterns in the CCA and CB. REFERENCES Batschelet E. Circular statistics in biology. London: Academic Press; 1981. Brown PM, Johnston KW, Kassam M, Cobbold RS. A critical study of ultrasound Doppler spectral analysis for detecting carotid disease. Ultrasound Med Biol 1982;5:515–523. Caro CG, Fitz-Gerald JM, Schroter RC. Arterial wall shear and distribution of early atheroma in man. Nature 1969;223:1159–1160. Castello R, Lenzen P, Aguirre F, Labovitz A. Variability in the quantitation of mitral regurgitation by Doppler color flow mapping: Comparison of transthoracic and transesophageal studies. J Am Coll Cardiol 1992;20:433–438.
Volume 40, Number 11, 2014 Evans DH, Jensen JA, Nielsen MB. Ultrasonic colour Doppler imaging. Interface Focus 2011;1:490–502. Fuster V, Stein B, Ambrose JA, Badimon L, Badimon JJ, Chesebro JH. Atherosclerotic plaque rupture and thrombosis. Evolving concepts. Circulation 1990;82(3 Suppl):II47–II59. Glagov S, Zarins CK, Giddens DP, Ku DN. Hemodynamics and atherosclerosis. Insights and perspectives gained from studies of human arteries. Arch Pathol Lab Med 1988;112:1018–1031. Hansen KL, Udesen J, Gran F, Jensen JA, Bachmann Nielsen M. In-vivo examples of flow patterns with the fast vector velocity ultrasound method. Ultraschall Med 2009a;30:471–477. Hansen KL, Udesen J, Oddershede N, Henze L, Thomsen C, Jensen JA, Nielsen MB. In vivo comparison of three ultrasound vector velocity techniques to MR phase contrast angiography. Ultrasonics 2009b; 49:659–667. Hansen KL, Udesen J, Thomsen C, Jensen JA, Nielsen MB. In vivo validation of a blood vector velocity estimator with MR angiography. IEEE Trans Ultrason Ferroelec Freq Contr 2009c;56:91–100. Hansen PM, Pedersen MM, Hansen KL, Nielsen MB, Jensen JA. New technology—demonstration of a vector velocity technique. Ultraschall Med 2011;32:213–215. Heijenbrok-Kal MH, Buskens E, Naderkoorn PJ, van der Graaf Y, Hunink MGM. Optimal peak systolic velocity threshold at duplex US for determining the need for carotid endarterectomy: A decision analytic approach. Radiology 2006;238:480–488. Hoit BD, Jones M, Eidbo EE, Elias W, Sahn DJ. Sources of variability for Doppler color flow mapping of regurgitant jets in an animal model of mitral regurgitation. J Am Coll Cardiol 1989;13: 1631–1636. Hugh AE, Fox JA. The precise localisation of atheroma and its association with stasis at the origin of the internal carotid artery—a radiographic investigation. Br J Radiol 1970;43:377–383. Jensen JA. Estimation of blood velocities using ultrasound: A signal processing approach. New York: Cambridge University Press; 1996. Jensen JA. A new estimator for vector velocity estimation. IEEE Trans Ultrason Ferroelec Freq Contr 2001;48:886–894. Jensen JA, Munk P. A new method for estimation of velocity vectors. IEEE Trans Ultrason Ferroelec Freq Contr 1998;45:837–851. Kassam MS, Cobbold RS, Johnston KW, Graham CM. Method for estimating the Doppler mean velocity waveform. Ultrasound Med Biol 1982;8:537–544. Munk P. Estimation of the 2-D flow vector in ultrasonic imaging: A new approach. Master’s thesis. Copenhagen: Technical University of Denmark, 1996. Murray CJ, Lopez AD. Alternative projections of mortality and disability by cause 1990-2020: Global burden of disease study. Lancet 1997;349:1498–1504. Pedersen MM, Pihl MJ, Haugaard P, Hansen JM, Hansen KL, Nielsen MB, Jensen JA. Comparison of real-time in vivo spectral and vector velocity estimation. Ultrasound Med Biol 2012;38:145–151. Peterson RE, Livingston KE, Escobar A. Development and distribution of gross atherosclerotic lesions at cervical carotid bifurcation. Neurology 1960;10:955–959. Phillips DJ, Greene FM, Langlois Y, Roederer GO, Strandness DE. Flow velocity patterns in the carotid bifurcations of young, presumed normal subjects. Ultrasound Med Biol 1983;9:39–49. Saba L, Sanfilippo R, Montisci R, Mallarini G. Correlation between USPSVand MDCTA in the quantification of carotid artery stenosis. Eur J Radiol 2010;74:99–103. Saba L. Imaging of the carotid artery. Atherosclerosis 2012;220:294–309. Sacco RL, Adams R. Albers. Guidelines for prevention of stroke in patients with ischemic stroke or transient ischemic attack. Stroke 2006; 37:577–617. Udesen J, Jensen JA. Investigation of transverse oscillation method. IEEE Trans Ultrason Ferroelec Freq Contr 2006;53:959–971. Udesen J, Nielsen MB, Nielsen KR, Jensen JA. Examples of in vivoblood vector velocity estimation. Ultrasound Med Biol 2007;33:541–548. Zarins CK, Giddens DP, Bharadvaj BK, Sottiurai VS, Mabon RF, Glagov S. Carotid bifurcation atherosclerosis. Quantitative correlation of plaque localization with flow velocity profiles and wall shear stress. Circ Res 1983;53:502–514.
