Optimization of ultrasound assessments of arterial function

doi:10.4236/ojcd.2011.13004

Paper Menu >>

Journal Menu >>

Open Journal of Clinical Diagnostics, 2011, 1, 15-21

doi:10.4236/ojcd.2011.13004 Published Online December 2011 (http://www.SciRP.org/journal/ojcd/

OJCD

Published Online December 2011 in SciRes. http://www.scirp.org/journal/OJCD

Optimization of ultrasound assessments of arterial function

Lee Stoner1,2*, Cary West 2, Danielle Morozewicz Cates2, Joanna M. Young3

1School of Sport and Exercise, Massey University, Wellington, New Zealand;

2Department of Kinesiology, University of Georgia, Ramsey Center, Athens, USA;

3Lipid and Diabetes Research Group, Diabetes Research Institute, Christchurch, New Zealand.

Email: *[email protected]

Received 16 September 2011; revised 18 November 2011; accepted 18 November 2011.

ABSTRACT

Ultrasound technology is widely used to make as-

sessments of arterial function. The delicate nature of

these measurements requires that sources of errors

are limited. Therefore, the aim of this study was to

assess variability due to probe selection and optimi-

zation settings. Methods: Ten healthy 20 - 26 year old

male and female subjects were tested. Brachial artery

size (diameter) was measured thirty times a second

using a B-mode Ultrasound unit equipped with a

high-resolution video capture device. Distension was

calculated using systolic and diastolic diameters. To

assess intersession variability, we made recordings

over twelve minutes; with the probe being removed

and re-positioned every four minutes. To assess vari-

ability due to probe selection and optimization, we

manipulated four parameters: 1) Probe selection (7 -

13 MHz, 5 - 10 MHz, 6 - 9 MHz); 2) Probe frequency

(11 MHZ, 9.6 MHZ, 8 MHz); 3) Measurement loca-

tion (near, center or middle field); And 4) Image

mode (B-mode, duplex-mode). To assess inter-session

variability, three sets of recordings were made for

each probe selection and optimization setting. Results:

Mean diameter ICC’s for inter-session variability,

probe frequency, measurement location, image dis-

play size, and probe selection were 0.99, 0.98, 0.97,

0.99, and 0.90 respectively. Distension ICC’s for

intersession variability, probe frequency, measure-

ment location, image display size, and probe selection

were 0.66, 0.26, 0.62, 0.60, and 0.51 respectively.

Conclusions: Altering probe selection increases

measurement variability to the greatest extent. How-

ever, as long as probe selection and optimization set-

tings are kept constant, our inter-session variability

shows that reliable measurements can be made.

Keywords: Ultrasound; Reproducibility; Diameters; Dis-

tension; Arterial Stiffness

1. INTRODUCTION

Ultrasound is widely used for the diagnostic assessment

of the carotid and peripheral arteries. The elastic proper-

ties of carotid and peripheral arteries are assessed by

studying dynamic properties of the arterial walls. Through

measurement of arterial distention, together with local

blood pressures, indices of arterial stiffness can be cal-

culated. Assessments of arterial stiffness have shown to

predict future cardiovascular complications [1-4].

In order to calculate arterial stiffness, the diameter of

a given artery must be continuously measured across the

cardiac cycle. For a carotid artery, this may entail meas-

urements that range from 8.0 mm to ~ 8.3 mm, a disten-

tion of 0.3 mm. For peripheral arteries, the distention

range will be much lower. Therefore, even small varia-

tions in systolic or diastolic diameters can notably im-

pact distention measurements. For this reason, it is im-

portant to limit possible sources of error. However, the

requirement for standardization of ultrasound technical

settings has not been reported in the literature.

The aim of this study was to assess variability due to

probe selection and optimization settings. To assess

variability due to probe selection and optimization four

parameters were manipulated: 1) Probe selection; 2)

Probe frequency; 3) Measurement location; And, 4) Im-

age mode. To assess inter-session variability, three re-

cordings were made for each parameter.

2. METHODS

2.1. Subjects

Ten healthy 20 - 26 year old male and female subjects

were tested. Informed consent was obtained from the

subjects after they were given a detailed description of

the procedures. The study was approved by the Univer-

sity of Georgia Institutional Review Board. Subjects

were excluded from the study if they demonstrated any

cardiovascular disease health risks or were taking medi-

cations with known vasoactive properties. Subjects were

L. Stoner et al. / Open Journal of Clinical Diagnostics 1 (2011) 15-21

asked to abstain from caffeine, high-fat foods, and alco-

hol for 24 hours prior to testing.

2.2. Protocol

Testing commenced following at least 20 minutes of

quiet supine rest. All measurements for a given subject

were made in one sitting. Brachial artery size (diameter)

was measured using a B-mode Ultrasound unit equipped

with a high-resolution video capture device. Diastolic,

systolic and mean diameters were recorded. Recordings

were made using eleven probe selection and optimiza-

tion settings (see Table 1). To assess inter-session vari-

ability, three sets of recordings were made for each

probe selection and optimization setting, with the probe

being removed and re-positioned every four minutes.

The three probes were linear array transducers. To com-

pare probes the highest imaging frequencies were set for

each probe (LA39, 11 MHz; 739, 9 MHz; 546, 6.6 MHz).

Aside from the probe comparison measurements, the

highest resolution probe (LA39) was used. Aside from

the location measurements, images were focused on the

center of the image display field. Aside from imaging

mode measurements, B-mode was used. Care was taken

to ensure that the same portion of the brachial artery was

imaged for all measurements. Subjects were asked to

hold their breath for ten seconds for each recording.

2.3. Diameter Measurements

High-resolution Brightness-mode (B-mode) ultrasound

measurements were made using a GE 400CL duplex

color Doppler unit (GE Medical, Milwaukee, Wisconsin).

The brachial artery of the left arm was measured in the

distal third of upper arm. Care was taken to ensure that

the vessel clearly extended across the entire [un-zoomed]

imaging plane to minimize the likelihood of skewing the

vessel walls. Magniﬁcation and focal zone settings were

then adjusted to optimize imaging of the proximal and

distal vessel walls. The image was comprised of 400 ×

400 pixels over an area of 16 × 16 mm, with a pixel

resolution of 0.04 × 0.04 mm. A specialized probe hold-

ing device enabled precise positioning and ensured that

pressure on the artery was minimized. The precise posi-

tion of the ultrasound probe was recorded and marked.

2.4. Diameter Analysis

Moving Picture Experts Group-2 (MPEG-2) recordings

were captured using a Dell Laptop PC equipped with a

video capture device (ADS technologies, Cerritos, Cali-

fornia). Video files collected at 30 frames/second were

converted to Joint Photographic Experts Group (JEPG)

images and subsequently used to make 30 diameter

measurements/second. JPEG images provide comparable

accuracy for ultrasound image measurements compared

Table 1. Probe and optimization setting parameters.

Optimization Setting Sub-Setting

LA39 (7 - 13 MHz)

739 (5 - 10 MHz)

Linear Array Probe

549 (6 - 9 MHz)

11 MHz

9.6 MHz

Probe Frequency

8 MHz

Near Field

Center field

Field Location

Far Field

B-Mode

Imaging Mode

PD-Mode (duplex)

to the Digital Image and Communications in Medicine

(DICOM) standard [5]. Images were measured offline

using semi-automated edge-detection software custom

written to interface with the LabVIEW data acquisition

platform (version 8.1, National Instruments, Austin,

Texas) [6,7]. Custom written Excel Visual Basic code

was used to fit peaks and troughs to diameter waveforms

in order to calculate diastolic, systolic, and mean diame-

ters. The within-session SEM3,1 for diameter measure-

ment with the described set-up is 0.046 mm. The be-

tween-day coefficient of variation is 2.7% for resting

diameter measurements [8].

2.5. Statistical Analysis

Statistical analysis was undertaken using SPSS 13 for

windows (SPSS Inc, Chicago, IL). The single measures

intra-class correlation coefficient (ICC) were calculated

using a two-way mixed effects (absolute agreement)

model where subject effects are random and the optimi-

zation/probe settings fixed. In general, values above 0.75

can be considered to represent excellent reliability, val-

ues between 0.4 and 0.75 represent fair to good reliabil-

ity and values below 0.4 represent poor reliability [9].

Standard error of measurement (SEM) and respective

confidence intervals were calculated using Eq. 1 and Eq.



SEMSD1 ICC (1)



95% CIMeanSEM1.96



(2)

where SD = the sample standard deviation, and ICC = as

calculated above.

Bland-Altman plots were constructed to provide an

indication of systematic bias and random error [10,11].

L. Stoner et al. / Open Journal of Clinical Diagnostics 1 (2011) 15-21 17

The 95% confidence intervals of limits of agreement

were calculated using Eq. 3:

95%1.96 SDd (3)

where: d = the sample bias (mean difference), and SD =

standard deviation of differences.

3. RESULTS

3.1. W ithin-Ses sion Va riability

Table 2 shows the inter-session variability for diameter

measurements. Single measure ICC values for Dm, Dd,

and Ds show excellent reliability. The %SEM is lowest

for Dd and highest for Ds. The Bland-Altman plot shown

in Figure 3 compares trials 1 and 3. There was no indi-

cation of systemic bias across the three trials. The ICC

for D∆ shows fair to good reliability. The %SEM is no-

tably higher for D∆ than for single diameters. Bland-

Altman plots for D∆ (not shown) do not indicate sys-

temic bias over the three trials.

3.2. Diameter Measurement Variability across

Ultrasound Settings

Table 3 shows the variability for diameter measure-

ments across ultrasound settings. Single measure ICC

values for Dm, Dd, and Ds diameters show excellent reli-

ability. The %SEM is lowest for Ds and highest for Dd.

The ICC for D∆ shows poor reliability. The %SEM is

notably higher for D∆ than for systolic or diastolic di-

ameters.

3.3. Diameter Measurement Variability for Each

Ultrasound Setting

Table 4 shows the variability for diameter measure-

ments for each ultrasound setting. Figures 1 and 2 show

example diameter analysis and resultant waveforms for

LA39 and 546 probes. ICC values for Dm, Dd, and Ds

across probe frequencies show excellent reliability.

Bland-Altman plots (not shown) show a bias for smaller

diameters as the probe frequency decreases.

The ICC for D∆ shows poor reliability. The %SEM is

notably higher for D∆ than for single diameters. The

Bland-Altman plot shown in Figure 4 compares fre-

quencies 11 MHz and 8 MHz for the LA39 probe. There

was no indication of systemic bias across the three probe

frequencies.

ICC values for Dm, Dd, and Ds across measurement

location show excellent reliability. Bland-Altman plots

(not shown) show a bias for smaller diameters as the

measurement location moves from the near- to the

far-side of the ultrasound imaging field. The ICC for D∆

shows fair to good reliability. The %SEM is notably

higher for D∆ than for single diameters. Bland-Altman

plots (not shown) show a bias for larger D∆ as the meas-

Figure 1. Examples of semi-automated diameter analysis on

images collected from (A) LA39 (11 MHz) and (B) 546 (6.6

MHz) probes. Images are captured at a rate of 30 images per

second. The images are identical to those shown in Figure 2.

A. LA39 (11MHz)B. 546 (6.6MHz)

3.7

3.8

3.9

4.0

0123

Time (s ecs)

Diamet er ( m m )

3.7

3.8

3.9

4.0

0123

Time (s ecs)

Diamet er ( m m )

Figure 2. Example of images collected from (A) LA39 (11

MHz) and (B) 546 (6.6 MHz) probes. Images were taken dur-

ing diastole and illustrate the first image used for analysis as

shown by the corresponding graphs above the images.

urement location moves from the near- to the far-side of

ultrasound imaging field.

ICC values for Dm, Dd, and Ds across imaging mode

show excellent reliability. Bland- Altman plots (not

shown) show a bias for smaller diameters when imaging

in PD versus B-mode. The ICC for D∆ shows fair to

L. Stoner et al. / Open Journal of Clinical Diagnostics 1 (2011) 15-21

OJCD

good reliability. The %SEM is notably higher for D∆

than for single diameters. Bland- Altman plots (not

shown) show no indication of bias.

ICC values for Dm, Dd, and Ds across probe selection

show excellent reliability. Compared to the other opti-

mization settings probe selection results in the lowest

ICC and highest %SEM. The Bland-Altman plot shown

in Figure 5 compares probes LA39 (11 MHz) and 539

(6.6 MHz). The 539 probe results in smaller diameters

than for LA39 and 739 (9 MHz) probes. The LA39 and

739 probes are more comparable. The ICC for D∆ shows

fair to good reliability. The %SEM is notably higher for

Table 2. Inter-session diameter measurement variability.

Dm Ds Dd D∆

T1 (mm) 4.143 (0.545) 4.189 (0.518) 4.076 (0.487) 0.113 (0.053)

T2 (mm) 4.149 (0.493) 4.196 (0.495) 4.076 (0.492) 0.120 (0.054)

T3 (mm) 4.148 (0.506) 4.198 (0.509) 4.081 (0.486) 0.117 (0.066)

Mean (mm) 4.147 (0.513) 4.194 (0.506) 4.078 (0.488) 0.117 (0.050)

ICC 0.992 0.991 0.996 0.661

SEM3,1 (mm) 0.046 0.048 0.031 0.029

SEM (%) 1.107 1.144 0.757 25.171

LCI (mm) 4.057 4.100 4.017 0.059

UCI (mm) 4.237 4.288 4.138 0.174

Diastolic, systolic, and mean diameters, and distension for each four-minute interval. Values are mean (SD).

Table 3. Diameter measurement variability across ultrasound settings.

Dm Ds Dd D∆

LA39 (mm) 4.121 (0.504) 4.161 (0.505) 4.089 (0.504) 0.072 (0.043)

739 (mm) 4.140 (0.539) 4.176 (0.542) 4.107 (0.530) 0.069 (0.046)

546 (mm) 4.040 (0.605) 4.101 (0.604) 4.002 (0.619) 0.099 (0.044)

11 mhz (mm) 4.183 (0.553) 4.252 (0.499) 4.168 (0.504) 0.084 (0.042)

9.6 mhz (mm) 4.127 (0.516) 4.191 (0.514) 4.090 (0.510) 0.101 (0.064)

8.2 mhz (mm) 4.109 (0.533) 4.148 (0.554) 4.070 (0.537) 0.078 (0.043)

Left (mm) 4.205 (0.487) 4.230 (0.502) 4.175 (0.482) 0.055 (0.050)

Cent. (mm) 4.179 (0.473) 4.209 (0.469) 4.145 (0.471) 0.064 (0.044)

Right (mm) 4.144 (0.516) 4.186 (0.527) 4.109 (0.510) 0.077 (0.047)

B (mm) 4.182 (0.476) 4.226 (0.477) 4.156 (0.481) 0.071 (0.042)

PD (mm) 4.151 (0.501) 4.191 (0.521) 4.127 (0.501) 0.064 (0.067)

Mean (mm) 4.144 (0.046) 4.190 (0.043) 4.113 (0.051) 0.077 (0.014)

ICC 0.915 0.918 0.914 0.372

SEM3,1 (mm) 0.013 0.012 0.015 0.011

SEM (%) 0.320 0.292 0.360 14.26

LCI (mm) 4.118 4.166 4.084 0.055

UCI (mm) 4.170 4.214 4.142 0.098

Diastolic, systolic, and mean diameters, and distension for each ultrasound setting. Values are mean (SD).

L. Stoner et al. / Open Journal of Clinical Diagnostics 1 (2011) 15-21 19

Table 4. Diameter measurement variability for each ultrasound setting.

Mean (mm) ICC SEM3,1 (mm) SEM (%) 95%CI (mm)

Dm 4.100 (0.532) 0.902 0.167 4.065 (3.774, 4.427)

Ds 4.146 (0.534) 0.909 0.161 3.885 (3.830, 4.462)

Dd 4.066 (0.534) 0.899 0.170 4.177 (3.733, 4.399)

Probe

D∆ 0.080 (0.037) 0.512 0.026 32.50 (0.029, 0.131)

Dm 4.140 (0.531) 0.980 0.075 1.815 (3.992, 4.287)

Ds 4.197 (0.519) 0.970 0.090 2.141 (4.021, 4.373)

Dd 4.109 (0.514) 0.975 0.081 1.980 (3.950, 4.269)

Freq.

D∆ 0.088 (0.036) 0.255 0.031 35.26 (0.027, 0.149)

Dm 4.176 (0.487) 0.967 0.088 2.118 (4.003, 4.349)

Ds 4.209 (0.495) 0.973 0.081 1.934 (4.049, 4.368)

Dd 4.143 (0.483) 0.966 0.089 2.148 (3.969, 4.318)

Location

D∆ 0.065 (0.041) 0.617 0.025 38.62 (0.016, 0.115)

Dm 4.166 (0.488) 0.994 0.038 0.908 (4.092, 4.240)

Ds 4.219 (0.497) 0.986 0.059 1.393 (4.103, 4.334)

Dd 4.145 (0.491) 0.995 0.035 0.838 (4.077, 4.213)

Display

D∆ 0.073 (0.051) 0.596 0.032 43.83 (0.010, 0.136)

Mean (SD) diameters for each ultrasound setting. Absolute difference (D), coefficients of variation (CV), intra-class cor-

relation coefficient (ICC), and standard error of measurement (SEM) plus confidence intervals (CI) are shown.

-0.20

-0.15

-0.10

-0.05

0.00

0.05

0.10

0.15

0.20

2.9 3.43.9 4.4 4.9

Average of T1 & T3 (m m )

T3 - T1 (mm )

Mean

- 1.96 S D

+ 1 .9 6

Figure 3. Bland-Altman plot of diameter difference values

(Trial 3 - Trial 1) on average diameter values ((Trial 1 + Trial

3)/2)). The mean difference, upper boundary ((mean + (SD of

mean × 1.96)), and lower boundary ((mean – (SD of mean ×

1.96)) are shown.

D∆ than for single diameters. Bland-Altman plots (not

shown) indicate bias towards larger D∆ for the 539 probe

compared to the other two probes.

4. DISCUSSION

This study shows that arterial diameters can be reliably

-0 .12

-0 .08

-0 .04

0.00

0.04

0.08

0.12

0.00 0.04 0.08 0.12

Avg. 11mhz & 8mhz ( m m )

Avg. 8m hz - 11m hz ( m m)

Mean

- 1.96 SD

+ 1 .96

Figure 4. Bland-Altman plot of distension difference values (8

MHz – 11 MHz) on average distension values ((11 MHz + 8

MHz)/2)). The mean difference, upper boundary ((mean + (SD

of mean × 1.9.6)), and lower boundary ((mean – (SD of mean

× 1.96)) are shown.

measured within-session, but that measurement error for

arterial distention calculation is notably higher. Also, it

was found that variations in ultrasound probe selection

and optimization settings contribute to some measure-

ment bias, though with little impact on the reliability of

diameter measurements. However, variations in ultra-

L. Stoner et al. / Open Journal of Clinical Diagnostics 1 (2011) 15-21

-0 .6

-0 .4

-0 .2

0.0

0.2

0.4

0.6

2.9 3.4 3.94.4 4.9

Avg. of LA39 & 546 ( m m )

Avg. 54 6 - LA39 (mm)

Mean

- 1.96 SD

+ 1.96 SD

Figure 5. Bland-Altman plot of diameter difference values

(546 – LA39) on average values ((LA39 + 546)/2)). The mean

difference, upper boundary ((mean + (SD of mean × 1.9.6)),

and lower boundary ((mean – (SD of mean × 1.96)) are shown.

sound probe selection and optimization settings notably

impact the reliability of arterial distention measure-

ments.

4.1. Probe Selection and Frequency Settings

Across ultrasound settings we found that probe selection

had the greatest impact on reliability for both diameter

and arterial distention measurements. The lower fre-

quency bandwidth probe (546, 6.6 MHz) resulted in bias

towards smaller diameter and distention measurements.

Decreasing the frequency of the LA39 probe also re-

sulted in bias towards smaller diameter values, although

this did not affect distension.

Probe selection and frequency settings may be de-

pendent on the population sample, i.e., subjects with

higher subcutaneous fat will require a probe that oper-

ates at a lower frequency bandwidth. Probe selection

may also be dependent on the artery being assessed, i.e.,

deeper arteries will require a lower frequency bandwidth.

Furthermore, during the course of an intervention study,

optimal probe selection may be dependent on body

compositional changes. This study shows that probe

selection needs to be standardized for a given subject if

repeated measures are to be made. Furthermore, diag-

nostic meaning for population comparisons may be in-

fluenced if different probes are used to compare popula-

tions.

4.2. Ultrasound Imaging Mode Selection and

Imaging Field

Another important finding of this study pertains to the

imaging display, i.e., B-mode vs. PD-mode (duplex). We

found that PD-mode results in marginally smaller di-

ameter and distention values. However, altering the im-

aging display did not affect the reliability of these meas-

urements. We found that changing the location from a

near-field to far-field did not notably impact reliability,

but did result in progressive bias towards smaller di-

ameters. The imaging display selected will be dependent

on whether simultaneous blood velocity measurements

are required. Reliable measurements can be made when

imaging in PD-mode, however if repeated measures are

to be made, then it is advisable that the imaging display

remain constant. It is also advisable that image focus is

maintained central to the ultrasound field.

5. CONCLUSIONS

Ultrasound can reliably measure arterial diameters and

distension for the duration of a given test session. How-

ever, alterations to probe selection and optimization set-

tings – particularly probe selection – can have a signify-

cant impact on measurement precision. Although reli-

ability is often more important than absolute accuracy

for serial exams, measurement differences due to varia-

tions in instrument settings may, nevertheless, be inter-

preted as having substantive diagnostic meaning. It is

therefore recommended that ultrasound probe selection

and optimization settings are standardized for repeated

measurements, and preferably across subjects.

REFERENCES

[1] Van Popele, N.M., et al. (2001) Association between

arterial stiffness and atherosclerosis: The Rotterdam

study. St roke, 32, 454-460. doi:10.1161/01.STR.32.2.454

[2] Cohn, J. (1999) Vascular wall function as a risk marker

for cardiovascular disease. Journal of Hypertension, 17,

S41-S44. doi:10.1016/S0895-7061(01)02154-9

[3] Cohn, J.N. (2001) Arterial compliance to stratify cardio-

vascular risk: More precision in therapeutic decision ma-

king. American Journal of Hypertension, 14,

S258-S263. doi:10.1016/S0895-7061(01)02154-9

[4] Kingwell, B.A. (2002) Large artery stiffness: Implica-

tions for exercise capacity and cardiovascular risk. Cli-

nical and Experimental Pharmacology and Physiology,

29, 214-217. doi:10.1046/j.1440-1681.2002.03622.x

[5] Hangiandreou, N.J., et al. (2002) The effects of irreversi-

ble JPEG compression on an automated algorithm for

measuring carotid artery intima-media thickness from ul-

trasound images. Journal of Digital Imaging, 15, 258-

260.

[6] Sabatier, M.J., et al. (2006) Doppler ultrasound assess-

ment of posterior tibial artery size in humans. Journal of

Clinical Ultrasound, 34, 223-230. doi:10.1002/jcu.20229

[7] Stoner, L., et al. (2006) Upper versus lower extremity

arterial function after spinal cord injury. Journal of Spi-

nal Cord Medicine, 29, 138-146.

[8] Stoner, L., et al. (2004) The relationship between blood

velocity and conduit artery diameter, and the effects of

smoking on vascular responsiveness. Journal of Applied

Physiology, 96, 2139-2145.

[9] Fleiss, J.L. (1986) The design and analysis of clinical

experiments. Wiley, New York.

[10] Bland, J.M. and Altman, D.G. (1999) Measuring agree-

L. Stoner et al. / Open Journal of Clinical Diagnostics 1 (2011) 15-21

OJCD

ment in method comparison studies. Statistical Methods

in Medical Research, 8, 135-160.

doi:10.1191/096228099673819272

[11] Bland, J.M. and Altman, D.G. (1986) Statistical methods

for assessing agreement between two methods of clinical

measurement. Lancet, 1, 307-310.

doi:10.1016/S0140-6736(86)90837-8