Key words vector flow imaging - vector concentration - velocity ratio - digital subtraction
angiography - peripheral arterial disease
Introduction
Peripheral arterial disease (PAD) has an age-adjusted prevalence of 4–15% and encompasses
a wide range of non-coronary arterial pathophysiological processes, which alter the
arterial supply to the brain, the visceral organs and the limbs [1 ]
[2 ]. Stenosis of the femoral artery is a disease entity within PAD, and is mainly caused
by atherosclerosis [1 ].
Contrast angiography provides detailed information about arterial anatomy and is recommended
for the evaluation of patients with stenosis of the femoral artery when revascularization
is contemplated. Digital subtraction angiography (DSA) is recommended for contrast
angiographic studies, because this technique allows enhanced imaging capabilities
compared with conventional unsubtracted contrast angiography and is considered the
“gold standard” for defining both normal vascular anatomy and vascular pathology in
PAD [1 ]. However, DSA is invasive, associated with risks of both local and systemic complications,
and exposes patients and staff to ionizing radiation [3 ].
An alternative to DSA is Doppler ultrasound (US) providing peak systolic velocity
estimation, velocity ratios within and beyond the stenosis, and evaluation of turbulence
[1 ]
[4 ]
[5 ]. Conventional Doppler US has a major limitation in terms of angle dependency, as
only the component of blood velocity directed along the axis of the emitted US beam
is measured. Therefore, assumptions of flow direction are necessary for flow quantification,
and flow complexity can only be evaluated in terms of flow towards and away from the
transducer [6 ].
The first solutions for angle independent US velocity estimation were proposed several
decades ago [7 ]
[8 ]
[9 ]
[10 ]. Later, a promising vector velocity method called transverse oscillation (TO) was
proposed by Jensen and Munk [11 ]. TO provides real-time, angle-independent blood flow estimation, and is currently
implemented in commercial scanners as the vector flow imaging (VFI) technique. The
TO method has been evaluated for flow estimation of various vessel geometries as reflected
in the most recent in-vivo studies [12 ]
[13 ]
[14 ]
[15 ]
[16 ]. However, the TO method has also been investigated for the estimation of cardiac
motion [17 ]
[18 ].
Velocity ratios obtained with TO have been used for stenosis assessment in the SFA
showing that TO can distinguish between stenoses over and under 50% lumen reduction
[19 ]. Another TO-derived parameter for stenosis assessment is vector concentration, a
measure of flow complexity, which showed a strong correlation to peak systolic velocities
obtained with transesophageal echocardiography (TEE) for flow changes induced by aortic
stenosis [20 ]
[21 ].
In this study, patients with stenoses in the SFA were examined with VFI and DSA. The
aim of the study was to investigate vector concentration and velocity ratio obtained
with TO compared with the stenosis degree percentage obtained with DSA for the assessment
of stenoses in the SFA in patients with chronic limb ischemia.
Materials and Methods
Patients
The TO and DSA data analyzed in this study are identical to the data used in a previous
comparison study of TO and DSA for SFA stenosis, where velocity ratios were calculated
from TO vector velocities [19 ]. Thirty consecutive patients with chronic limb ischemia scheduled for endovascular
therapy of the lower extremities were included. Patients were eligible for inclusion
if they had one or more previously untreated arteriosclerotic lesions in the SFA.
Nineteen patients were excluded due to previous bypass surgery, endovascular surgery,
occlusion, no lesions (judged by both US and DSA), or widespread atherosclerotic disease
according to The Trans-Atlantic Inter-Society Consensus Document on Management of
Peripheral Arterial Disease [22 ]. Eleven patients were included with a total of 16 lesions (7 males, 4 females, mean
age: 71.6 years, range: 53–84 years). Written informed consent was obtained from all
patients. The local ethics committee waived approval, since US scanning of extremities
with PAD is considered a routine procedure (no: H-4–2013–001).
Scan setup
A commercial scanner (UltraView 800, BK Medical, Herlev, Denmark) equipped with a
linear transducer using a center frequency of 9 MHz (8670, BK Medical, Herlev, Denmark)
was used for the US examinations. All patients were scanned in the angio-suite after
15 min of rest prior to DSA. The US scan included examination of the bifurcation of
the common femoral artery down to the level of the SFA, where it enters the adductor
canal. Turbulent/disturbed flow was detected in the long-axis view with TO, where
vortices and/or sudden aliasing occurred indicating increasing blood flow velocity.
The lesion was centered in the scan plane, so flow was estimated within, proximal
and distal to the lesion in the same recording ([Fig. 1 ]). A radiodense marker was attached to the skin corresponding to the anatomic location
of the TO scan. In the subsequent DSA, the marker pointed directly towards the suspected
lesion, ensuring matching ultrasonic and angiographic recordings.
Fig. 1 Vector velocity images of 2 patients with stenosis of the SFA are shown in a and b , where A corresponds to lesion no. 3 and B to lesion no. 1 ([Table 1 ]). Both frames are taken from systole. The lesions are marked with an asterisk, and
in each frame, the ROI for calculation of the vector concentration is illustrated
with a white box. Direction and velocity of the blood flow estimated with TO are shown
by the color map.
VFI and calculation of vector concentration
The VFI method estimates both the axial and transverse velocity component using emissions
of conventional pulses for Doppler ultrasound. The motion in the axial direction is
found as in conventional Doppler ultrasound, while the motion in the transverse direction
is found by using a changed apodization in receive beamforming and a special estimator
[23 ]. Previous papers provide detailed explanations of the TO method [11 ]
[23 ]
[24 ].
The VFI color box was adjusted to cover the vessel with the lesion along with a disease-free
adjacent vessel segment included. The adjacent disease-free vessel segment was defined
as a segment of the SFA with no narrowing of the lumen and with laminar flow, i. e.,
without vortices and/or sudden aliasing. The pulse repetition frequency (PRF) was
adjusted for each scan ensuring optimal filling of the vessel in both the stenosed
and the adjacent disease-free vessel segment, even if aliasing in systole occurred.
Wall filter and color gain were likewise adjusted to the level providing optimal filling
of the vessel without flow artifacts outside the vessel. All other settings remained
in default mode. The angle of insonation was 70–90 degrees in all cases. The temporal
resolution of the TO estimation was 16 frames/s, and the maximum scan depth was approximately
5 cm due to the available transducer setup. The recorded TO cine loop of 14 s corresponded
to 225 frames. An overview of the applied scan settings along with TO acquisition
parameters is given in [Table 1 ]. On the US scanner, vector velocity estimates were displayed in real-time, but without
any quantification of velocities available. Afterwards, the cine loops were analyzed
off-line using MATLAB (MathWorks, Natick, MA, USA) as previously described [25 ].
Table 1 TO acquisition setup and scan settings with standard deviation (SD) in parentheses.
Number of elements
128
Bandwidth
70%
Pitch
0.3 mm
Kerf
0.035 mm
Height
4 mm
Elevation focus
20 mm
Pulse length
6 cycles sinusoidal
Lateral wavelength
4 * pitch
Average PRF
3.3 kHz (1.6 kHz)
Average wall filter cutoff frequency
134.8 Hz (73.1 Hz)
Average gain
52.3% (3.6%)
The averaged vector concentration r of the blood flow in the SFA during 5 consecutive systoles from the beginning of
the recorded cine loop was calculated for each examined vessel segment. The region
of interest (ROI) used for vector concentration estimation included the lesion along
with the disturbed flow in the periphery of the lesion ([Fig. 1 ]). The vector concentration is a calculation of the vector angle spread within the
ROI, and a measure of flow complexity [26 ]. In brief, the vector concentration is calculated as follows. For each position
i in the vector map, where the axial velocity component v
x
and the transverse velocity component v
z
are estimated, the flow angle θ
i
of the vector is:
Each flow angle θ
i
is represented on the unit circle as P
i
=(x
i
, y
i
), where x
i
=cos(θ
i
) and y
i
=sin(θ
i
). For each ROI encompassing the entire vessel, the mean value for x
i
and y
i
is found as,
To quantify the flow complexity, i. e., the vector angle spread, the vector concentration
r is found using Pythagoras’ theorem:
where r is one for perfect laminar flow, and decreases towards zero with increased complex
flow. Thus, vector concentration can be regarded as an in-vivo measure of flow complexity comparable to the Reynolds number, which is used to predict
flow patterns in fluid mechanics [27 ].
Furthermore, from each TO recording, 3 frames illustrating flow with the best possible
filling of the vessel in both the lesioned and the healthy part of the SFA were selected.
The velocity ratio calculated from TO estimates of each stenosis was found as the
maximum velocity detected centrally in the lesioned segment divided by the maximum
velocity detected centrally in the adjacent disease-free segment. The frames were
obtained from the cardiac cycle, where flow in both the stenosed and adjacent disease-free
vessel segment was antegrade and without aliasing. The velocity ratios were not calculated
at identical time points in the cardiac cycle. However, a constant velocity ratio
for each stenosis was assumed, when the velocities used for the calculation of the
velocity ratio were acquired from the same frame, i. e., at the same time point [19 ]. The velocity ratios reported in this paper are identical to the velocity ratios
found in a previous paper, where a more thorough explanation of the velocity ratio
estimation is given [19 ].
The calculations of vector concentration and velocity ratio were performed separately
and by 2 different experienced radiologists (KLH and PMH) blinded to the corresponding
results of the DSA.
Angiography
An Infinix-i system (model INFX-8000V, Toshiba Medical Systems Corporation, Tochigi-ken,
Japan) was used for DSA. An 11-cm 5-Fr sheath was placed in the artery. A 4- or 5-Fr
catheter was used when needed for contrast injections. DSA of the femoral artery was
performed using 2 frames/s and a 6–10 ml iodine contrast injection (Visipaque 270 mgI/ml,
GE Healthcare). Routine anteroposterior images in one plane were recorded and occasionally
supplemented by oblique projections. Subsequent measurements were performed on a standard
workstation. The DSA image yielding the most severe diameter reduction was used for
calculation of the stenosis degree percentage, i. e., the smallest diameter in the
stenosis vs. the diameter in an adjacent normal arterial segment. An experienced radiologist
not otherwise involved in this study and blinded to the result of the corresponding
US scan, calculated the stenosis degree percentage for each patient.
Statistical analysis
Measurements obtained with TO were initially analyzed with descriptive statistics,
i. e., mean and standard deviation (SD). The TO measurements, i. e., vector concentration
and velocity ratio, were then compared with the stenosis degree percentage obtained
with DSA using linear regression analyses with a 2-tailed significance value and p <0.05 considered significant. The correlation coefficient, regression equation, and
confidence interval (CI) using Fisher’s r -to-z -transformation were calculated. Statistical analyses were performed with IBM SPSS
Statistics v. 19 (SPSS Inc., Chicago, IL, USA).
Results
The descriptive statistics on vector concentration and velocity ratio with the corresponding
stenosis degree percentage are given in [Table 2 ]. The mean vector concentration and the mean velocity ratio obtained with TO were
0.67 (SD: 0.21) and 1.25 (SD: 0.35), respectively, and the mean stenosis degree percentage
obtained with DSA was 33.6% (SD: 27.8%). The association between vector concentration
and stenosis degree percentage, and between velocity ratio and stenosis degree percentage
are illustrated with scatterplots in [Fig. 2 ]. For the comparison between vector concentration and DSA, the correlation coefficient
R was 0.93 (p <0.001; 95% CI: 0.81 to 0.98), and the regression equation was y =−0.007x +0.917, where x corresponded to the stenosis degree percentage and y to the vector concentration. For the comparison between velocity ratio and DSA, the
correlation coefficient R was 0.50 (p <0.07; 95% CI: 0.00 to 0.80), and the regression equation was y =−0.012x +1.210, where x corresponded to the stenosis degree percentage and y to the velocity ratio. The correlation coefficients for the 2 comparison analyses
were significantly different (p <0.005) without overlapping CI.
Fig. 2 Scatterplots of TO-derived vector concentration and velocity ratio compared with
DSA-derived stenosis degree percentage. Line of best fit is illustrated with a black
solid line for each subplot.
Table 2 Averaged vector concentration and velocity ratio with standard deviation (SD) in
parentheses along with the corresponding DSA-derived stenosis degree percentage for
each stenosis examined.
Patient no.
Lesion no.
Shadowing calcifications (yes/no)
Velocity ratio (SD)
Vector concentration (SD)
Stenosis degree percentage [%]
1
1
N
2.2 (0.41)
0.35 (0.04)
78
1
2
Y
1.1 (0.15)
0.94 (0.03)
0
2
3
N
1.0 (0.06)
0.96 (0.03)
0
2
4
Y
1.2 (0.06)
0.73 (0.06)
19
3
5
N
2.9 (0.58)
0.41 (0.06)
68
4
6
N
2.6 (1.59)
0.57 (0.04)
65
5
7
Y
1.2 (0.15)
0.79 (0.03)
37
5
8
Y
0.9 (0.32)
0.81 (0.05)
31
6
9
Y
2.1 (0.30)
0.66 (0.01)
33
6
10
Y
1.5 (0.15)
0.82 (0.09)
15
6
11
Y
1.2 (0.15)
0.75 (0.13)
15
7
12
N
2.2 (0.17)
0.40 (0.07)
62
8
13
Y
2.5 (0.55)
0.71 (0.08)
11
9
14
N
1.2 (0.10)
0.48 (0.17)
47
10
15
N
1.3 (0.06)
0.95 (0.02)
0
11
16
N
1.0 (0.06)
0.38 (0.09)
67
Discussion
This study of stenosis assessment in the SFA indicated that vector concentration was
more strongly associated than velocity ratio to stenosis degree percentage as the
R -value was higher without overlapping CI for corresponding correlation analyses. The
presented results support previous studies of vector concentration showing excellent
performance in the evaluation of aortic valve stenosis with intraoperative VFI of
blood flow in the ascending aorta [20 ]
[21 ]
[28 ]. The 2 most recent studies showed that the flow complexity quantified with vector
concentration was different among patients with a normal, stenosed, and replaced aortic
valve (p <0.0001), and with a strong association to peak systolic velocity (p <0.0001, R =0.87 and 0.88) [20 ]
[21 ].
Flow complexity can be assessed in conventional US using estimation of spectral broadening
in spectral Doppler, power intensity in power Doppler, or by evaluation of mosaic
patterns using color Doppler [29 ]
[30 ]
[31 ]. However, stenosis grading in conventional US is normally done by velocity estimation,
such as measurements of peak velocities and mean gradients in aortic valve stenosis,
and velocity ratios in stenosis of the SFA [5 ]
[32 ].
The TO data in this paper have previously been used for velocity ratio estimation
showing that TO was able to separate stenosis over and under 50% lumen reduction (p <0.01). Two patients were considered outliers (patient 8 and 11) [19 ]. To test performance, linear correlation analysis was done for both velocity ratio
and vector concentration compared with DSA in this study. So, applying the same TO
data for evaluation of vector concentration instead of velocity ratio gave an improved
association to the same data set of DSA-derived stenosis degree percentage, and without
any outliers. Therefore, vector concentration may be a better parameter for the evaluation
of stenosis and the resultant flow changes in the SFA than velocity ratio estimation
when using vector velocity data.
There are several reasons for this improvement. TO is based on conventional pulsed
US emission, and therefore is limited by aliasing like conventional Doppler US. However,
vector concentration estimation is less PRF-sensitive than velocity estimation. While
aliased laminar blood flow will result in velocity estimates with large errors, the
flow direction will be reversed but uniform, and therefore, have a less affected vector
concentration [21 ]
[28 ]. As previously stated, some reduction in the vector concentration estimate has to
be expected when aliasing occurs even for laminar flow, as the central part of the
flow with the highest and aliased velocities will appear retrograde, whereas the peripheral
flow will remain antegrade [21 ]. Thus, at least 2 opposing flow angles within the ROI will be present, which will
increase the vector angle spread of the ROI, and in the case with an equal amount
of scatter moving in 2 opposite directions, the vector concentration will approach
zero. Nevertheless, this effect is less prominent than the effect of the actual increase
in flow complexity for increasing vessel stenosis. This has been indicated in previous
papers, where TO-derived vector concentration estimates compared with peak systolic
velocities obtained with continuous wave US using TEE had a strong linear relationship,
even when including vector concentration estimates obtained above the Nyquist limit
[20 ]
[21 ]. In the previous papers evaluating vector concentration for aortic valve stenosis,
the PRF was adjusted to the peak systolic velocity to reduce aliasing, while the PRF
in this study was adjusted to optimal filling of the lesioned and adjacent disease-free
vessel segment even if aliasing occurred. This underlines the need for a more thorough
evaluation of vector concentration in relation to PRF settings.
Another advantage of vector concentration estimation is less dependency on peak flow
alignment. The blood flow in-vivo is often non-parabolic, and therefore, the highest velocities are not always found
in the center of the lumen but along the vessel wall as shown in a previous study
with VFI and MRI of flow in the carotid artery [33 ]. Also, a VFI study of blood flow in the ascending aorta showed only a moderate association
with TEE for peak systolic velocity estimation, and a weak association with pulmonary
artery catheter thermodilution for cardiac output estimation, mainly due to asymmetry
of the aortic flow [34 ]. In vector concentration estimation, only the direction of flow is used. If the
blood flow is laminar, it will be uniform even for off-center evaluation, and likewise,
if the blood flow is complex, it will behave complex for off-center evaluation. Thus,
it is not crucial to capture the jet of the flow, as all blood flow layers are representative
of the flow complexity reflecting the degree of stenosis.
The angle-independent velocity estimation in TO is an obvious improvement compared
with blood flow estimation in conventional US systems, as flow direction and velocity
to every pixel within an ROI are given. In vector concentration estimation, all flow
data within the ROI are used providing much more data for the evaluation when compared
with the velocity estimation, whether obtained with conventional Doppler US or VFI.
The reduced dependency of flow alignment and the use of more flow data are an advantage
when stenoses are calcified creating acoustic shadowing. A previous study of carotid
artery stenosis showed that in the presence of calcifications, spectral Doppler US
is inadequate to accurately determine the degree of stenosis when compared with computed
tomographic angiography [35 ]. According to [Table 2 ], 50% of the included patients in this study had shadowing calcifications in the
SFA stenosis, and vector concentration still correlated well with the corresponding
stenosis degree percentage. Moreover, the 2 patients (8 and 11) acting as outliers
for velocity ratio estimation both had calcifications ([Table 2 ]).
VFI estimation is probably less user-dependent and more robust than default velocity
estimation as angle correction and range gate positioning are avoided. Two recent
papers indicated that VFI is more precise and accurate and less angle- and operator-dependent
than conventional spectral Doppler US for in-vivo velocity estimation [12 ]
[13 ]. With the introduction of vector concentration, VFI may also prove to be less PRF-dependent
as indicated by this and previous papers [20 ]
[21 ]
[28 ].
This study had several limitations. First of all, the study population was small.
Also, vector concentration was not calculated directly on the scanner but processed
off-line. However, the calculations were not computationally demanding, and should,
therefore, be easy to implement on the commercial platform for real-time estimation.
The vector concentration estimation could be further improved with a more automated
and less user-dependent interface, where predefined scan and ROI settings guide the
user. Only one operator performed the examinations in this study, hence, no interobserver
variability was assessed. Inter- and intraobserver variability of velocity estimation
has been investigated for VFI and revealed superior performance compared with spectral
Doppler US [13 ]. DSA is a 2D method, and underestimation of stenoses can therefore occur if the
smallest diameter of the vessel is not shown in the angiographic projection. DSA is
occasionally supplemented by oblique projections if any doubt about a stenosis is
raised, but that is no guarantee for a projection illustrating the most severe stenosis
degree. In [Fig. 1a ], a stenosis shown with US corresponded to a stenosis degree percentage of 0% according
to DSA ([Table 2 ]) illustrating this limitation.
In this study, spectral Doppler was not used as a reference, as DSA is considered
the gold standard for diagnosing and grading PAD. However, in future studies of vector
concentration obtained with VFI, spectral Doppler will be employed as a reference
method for a more complete validation. The optimal parameter settings for vector concentration
estimation, e. g., PRF, gain and wall filter, should be investigated in a controlled
setup, and flow disturbances in other vessel geometries, e. g., stenosis of the carotid
artery, in arteriovenous fistula, and of the valves of the heart using vector concentration
should also be explored. The studies should include larger patient populations and
more US operators. The TO-derived vector concentration could be a potential powerful
clinical tool for a reliable, easy-to-use, fast, and non-ionizing method of stenosis
assessment in patients suffering cardiovascular diseases such as PAD.
Conclusion
Vector concentration is a new parameter for blood flow evaluation obtained with angle-independent
vector velocity estimation. In this study of PAD in the SFA, vector concentration
was superior to velocity ratio for stenosis assessment when compared with stenosis
degree percentage obtained with DSA. The study indicates that vector concentration
could be a better parameter for stenosis evaluation than evaluation of blood flow
velocities when using VFI.
