- Email: [email protected]
Green Tagging in Displaying Color Doppler Aliasing: A Comparison to Standard Color Mapping in Renal Artery Stenosis
Ultrasound in Med. & Biol., Vol. 39, No. 11, pp. 1976–1982, 2013 Copyright Ó 2013 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.2013.05.006
d
Original Contribution GREEN TAGGING IN DISPLAYING COLOR DOPPLER ALIASING: A COMPARISON TO STANDARD COLOR MAPPING IN RENAL ARTERY STENOSIS JING GAO,* KEVIN MENNITT,* LILY BELFI,* YUAN-YI ZHENG,y ZONG CHEN,z and JONATHAN M. RUBINx * Department of Radiology, New York-Presbyterian Hospital, Weill Cornell Medical College, New York, New York, USA; y Institute of Ultrasound Imaging, The Second Affiliated Hospital, Chongqing Medical University, Chongqing, China; z Department of Computer Science, Fairleigh Dickenson University, Teaneck, New Jersey, USA; and x Department of Radiology, University of Michigan, Ann Arbor, Michigan, USA (Received 12 December 2012; revised 7 May 2013; in final form 14 May 2013)
Abstract—To quantitatively assess the contrast-to-noise ratio (CNR) of green tagging and standard color flow images in displaying fast flow velocity, we retrospectively reviewed 20 cases of hemodynamically significant renal artery stenosis (RAS) detected by renal color Doppler ultrasound and confirmed with digital subtraction angiography. At the site of RAS, blood flow with high velocity that appeared as aliasing on color flow images was computationally analyzed with both green tagging and standard color mapping. To assess the difference in the CNR between normal background flow and the aliased signal as a function of visualizing aliasing between the two color mappings, we used GetColorpixels (Chongqing Medical University, Chongqing, China) to count the values in the color channels after segmenting color pixels from gray-scale pixels. We then calculated the CNR in each color channel—red, green, and blue (RGB)—in the aliasing region on green tagging and standard color mapping. The CNRs in the red, green and blue channels were 0.35 ± 0.44, 1.11 ± 0.41 and 0.51 ± 0.19, respectively, on standard color mapping, and 0.97 ± 0.80, 4.01 ± 1.36 and 0.64 ± 0.29, respectively, on green tagging. We used a singlefactor analysis of variance and two-tailed t-test to assess the difference in CNR in each color channel between the two color mappings at the site of RAS. With these comparisons, there was no significant difference in the CNR in the red or blue channel between green tagging and standard color mapping (p . 0.05). However, there was a statistically significant difference in the CNR in the green channel between the two color mappings (p 5 0.00019). Furthermore, the CNR measured in the green channel on the green tagging image was significantly higher than the CNRs in all other color channels on both color mapping images (p 5 0.000). Hence, we conclude that green tagging has significantly higher visibility as a function of high-velocity flow than standard color mapping. The use of green tagging may improve the ability to detect RAS with color Doppler ultrasound. (E-mail: [email protected]) Ó 2013 World Federation for Ultrasound in Medicine & Biology. Key Words: Color Doppler ultrasound, Contrast-to-noise ratio, Green tagging, Renal artery stenosis.
represents local high velocities that can be used to identify areas of potential stenosis. Thus, given the proper settings for color Doppler frequency and other imaging parameters (total color gain, PRF, wall filter) (Utsunomiya et al. 1990), the identification of local high velocities strongly suggests a reduction of the cross-sectional area of the artery being imaged, that is, arterial luminal narrowing. The velocities increase as a result of renal auto-regulation to maintain constantvolume flow through the narrowed vessel (Rubin 1995). Therefore, the site of aliasing, as a function of highvelocity flow, is suspicious for RAS. Most importantly, the site of aliasing is highly suggestive of an optimal location for placing a spectral Doppler range gate for measuring the local Doppler spectrum.
INTRODUCTION Color Doppler ultrasound (CDUS) is commonly used in screening for renal artery stenosis (RAS), a condition characterized by clinically uncontrolled hypertension and/or renal failure (Lee and Grant 2002; Spyridopoulos et al. 2010). During renal CDUS, color flow imaging is used to assess mean flow velocity in renal vessels. When the pulse repetition frequency (PRF) is appropriately set for blood flow in the kidney and renal artery, local aliasing on color flow imaging Address correspondence to: Yuan-Yi Zheng, Institute of Ultrasound Imaging, The Second Affiliated Hospital, Chongqing Medical University, 76 Lijiang Road, Yuzhong District, Chingqing, China 400010. E-mail: [email protected] 1976
Green tagging vs. standard color mapping in renal artery stenosis d J. GAO et al.
The color flow image is not limited to display of mean velocity. Additional information, such as the spread of velocities within the sampled region, can also be depicted. Therefore, variance color mapping (commonly coded green, hence green tagging) is specifically designed to improve the sensitivity of displaying the variation in Doppler shifts within a sampling region, as a function of Doppler velocity at the region of interest. Thus, in a variance map, the local intensity of green in the 2-D color Doppler image represents a statistical measure of the spread of the velocities around each local mean velocity estimate (Hedrick and Hykes 1992). This is particularly useful in distinguishing areas of turbulent flow, and it was initially used in echocardiography to improve the detection of minimal valve regurgitation jets (Switzer et al. 1987). Green tagging has also been used in vascular sonography to delineate carotid artery stenosis (Baxter and Polak 1994). Because green tagging has been found to be of great value in displaying Doppler shift changes in cardiovascular ultrasound, we hypothesized that the contrast-tonoise ratio (CNR) of green tagging would be better than that of standard color Doppler mapping in detecting fast flow velocity at the site of RAS. The ultimate goal of this comparative study was to evaluate which color mapping has a greater ability to detect RAS with CDUS. METHODS This retrospective study was approved by the institutional review board (IRB 1201012129) of Weill Cornell Medical College and conducted at the New YorkPresbyterian Hospital, New York. The requirement for informed consent was waived. The study was compliant with the Health Insurance Portability and Accountability Act of 1996. Patients From January 2010 to June 2011, 20 patients with RAS (12 men and 8 women, mean age 5 59 6 14 y, range: 25–80 y) were identified by renal CDUS, using both green tagging and standard color mapping, and confirmed with digital subtraction angiography (DSA). All 20 cases with hemodynamically significant RAS (.60% arterial lumen narrowing, 10 with native RAS and 10 with transplant RAS) had a medical history of uncontrolled hypertension (rest blood pressure .150/90 mm Hg, n 5 11), renal failure (serum creatinine .1.1 mg/dL, n 5 4) or both (n 5 5). We reviewed static images and cine loops of renal CDUS recordings and static images of the DSA through a web-based Picture Archiving and Communication System (PACS, Centricity Enterprise Web V2.1, GE Medical Systems, Milwaukee, WI, USA). We searched our institutional electronic medical records system for
1977
pertinent subject demographic information including sex, age and reason for undergoing renal CDUS and DSA. We included 20 cases with RAS detected with both standard color mapping and green tagging with the same color Doppler parameters. All cases of RAS were confirmed with DSA. Renal color Doppler sonography No specific preparation for renal transplant CDUS was required before examinations. All patients who underwent native renal CDUS were requested to fast for 6–8 h before their ultrasound examinations. Patients were placed in the supine position for renal transplant CDUS and left or right decubitus position for native renal CDUS. A curved linear array transducer (4C1) or sector transducer (4V1) with multi-frequencies of 2–4 MHz (Acuson Sequoia 512 or S2000, Siemens Medical Solutions, Mountain View, CA, USA) was used to acquire color Doppler images. All scans were performed according to the American Institute of Ultrasound in Medicine (2009) practice guideline for the performance of renal artery duplex sonography. We started renal sonography by measuring renal size, observing cortical echogenicity and evaluating renal hydronephrosis, calculus or mass on gray-scale imaging. Color flow imaging was then performed to assess renal vasculature. Color Doppler settings including PRF, wall filter and overall color gain were adjusted from minimum to maximum to observe arteries in the renal parenchyma with slow flow and to visualize renovascular abnormalities with fast flow, that is, intra-renal RAS. Detection of vascular abnormalities in the kidney is highly dependent on the color Doppler settings. A low PRF may cause Doppler aliasing that overlaps high-velocity flow, resulting in a decrease in the distinction between aliased fast flow velocity at the stenotic artery and slow velocity flow in a normal artery. On the other hand, a high PRF could suppress slow flow in intra-renal vessels to make fast flow velocities at the stenosis obvious (Gao et al. 2010). In our institution, color Doppler was pre-set with standard color mapping in all ultrasound scanners. Green tagging was manually selected for recognizing fast flow velocities when the aliasing was not recognizable on standard color mapping images in the cases highly suspicious for RAS. It is known that machine parameters strongly affect the display of fast flow velocities as aliasing in color flow imaging (Utsunomiya et al. 1990). To minimize variations in color values produced by different color Doppler parameter settings and ultrasound machine design, we only included 20 cases of RAS imaging acquired using both standard color mapping (Fig. 1a) and green tagging (Fig. 2a) with the same color Doppler parameters on the same ultrasound scanner.
1978
Ultrasound in Medicine and Biology
Volume 39, Number 11, 2013
Fig. 1. Renal color Doppler sonography was performed on a renal transplant patient who developed uncontrolled hypertension. The arterial anastomosis, transplanted renal artery and iliac artery were displayed on a zoomed color Doppler image with standard color mapping (a). The rectangular box is the region of interest (ROI), in which questionable color Doppler aliasing appears at the site of a suspected arterial stenosis (a). Also shown is the zoomed ROI with color segmentation (b) processed with software (GetColorpixels). To calculate the contrast-to-noise ratio of the standard color mapping image, color pixels in the ROI with suspicious high velocity and the background (entire color box excluding ROI) are counted.
A spectral Doppler gate of 2–4 mm was placed at the center of the artery to measure the maximum velocity with Doppler angle correction by adjusting the emission sound beam to be as parallel as possible to the flow direction (at least ,60 ). The spectral Doppler waveforms were recorded for flow velocity measurements. We also assessed the abdominal aorta (in native renal sonography) and the iliac artery (in renal transplant sonography) with color flow imaging and spectral Doppler. Peak systolic velocity (PSV) was manually measured at the renal artery (proximal, mid and distal portions), abdominal aorta and iliac artery with electronic calipers. Special attention was paid to an artery with focal aliasing throughout the scan. The ratio of the PSV of the stenotic artery to that of the
artery proximal to the stenosis was calculated. A diagnosis of RAS was considered when the renal artery/aorta (iliac artery in renal transplant) PSV ratio was greater than 3.5 and the PSV at the renal artery was .1.8 m/s in native kidney (.2.6 m/s in transplanted kidney) (Gao et al. 2010). All 20 cases with RAS were confirmed with DSA. Quantification of color Doppler signal To count pixels in each color channel, color pixels must be separated from gray-scale pixels (Borchers and Klews 1993). For standard color mapping, the colorcoded image representing blood flow direction and mean velocity is overlaid on the gray-scale (B-mode)
Fig. 2. With the same color Doppler priority settings as used for standard color mapping in Figure 1, green tagging (green arrow) was applied to color flow images at the same anatomic location as in Figure 1. (Note that the red/blue color distributions slightly differ between Figs. 1 and 2.) It is clearly noted that significant aliasing appears as green at the proximal transplanted renal artery. The rectangular box contains the region of interest (ROI) (a). A high flow velocity appears in the ROI with color segmentation (b). The background for calculating the contrast-to-noise ratio is the entire color box excluding the ROI.
Green tagging vs. standard color mapping in renal artery stenosis d J. GAO et al.
image to form an output digital color Doppler image display on the screen of an ultrasound scanner. Moreover, fast flow with a wide range of flow velocities was tagged as green that is overlaid on a standard color mapping image to form a green tagging image. Hence, green tagging is considered a velocity tagging function that emphasizes a selected range of velocities with a high variance by highlighting regions of flow containing such velocities in green. On the ultrasound scanner, the color bar adjacent to the displayed image indicates in green (Fig. 2a, green arrow) that portion of the velocity scale selected for tagging. For offline image processing, de-identified (anonymized) static images of green tagging and standard color mapping with TIFF format were transferred to a personal computer. Ultrasound image analysis software (GetColorpixels, Chongqing Medical University, Chongqing, China) was developed to quantify the color pixels. Color pixels representing color Doppler signals need to be separated from the background gray pixels in color Doppler images. To avoid mis-registration during this segmentation, a threshold value was required to adequately distinguish between gray and color pixels (Jun et al. 2002). In a standard Doppler image, the background is mapped with pixels with little variance in R, G, B components, whereas the Doppler signals were mapped with pixels with large differences in R, G, B components, which could be used to differentiate the different pixels by setting a threshold. If the value (maximum value of R/G/B – minimum value of R/G/B) is smaller than this threshold, the pixel is considered a background pixel; otherwise, it is considered a Doppler signal pixel. We set a threshold of 50, which completely deleted the gray pixels (Figs. 1b and 2b), allowing us to segment the color pixels (Jun et al. 2002). After segmenting color pixels, we used the Image J (National Institutes of Health, Bethesda, MD, USA) histogram to calculate the pixels in the region that was previously gray. In all cases, a zero value was measured, which indicated that all gray-scale pixels had been completely deleted, leaving only color pixels in each image. It was now possible to count the segmented color pixels. This threshold may delete some color pixels during the processing; however, as long as we used the same threshold setting for all cases, there should be no effect on the comparison of color pixel values in each color channel between standard color mapping and green tagging. Finally, the color values in each color channel, R, G and B, was measured with the software GetColorpixels at the region of interest and background to calculate the CNR. Contrast-to-noise ratio of green tagging and standard color mapping The CNR is a standard method for estimating the ability to detect targets in diagnostic imaging (Song
1979
et al. 2004; Techavipoo and Varghese 2005). In screening for RAS, color flow imaging is used to detect the location of aliasing as a function of high-velocity flow at the site of arterial stenosis, which is further evaluated with spectral Doppler velocity analysis. Therefore, it is important to recognize local aliasing on color Doppler images. Region of interest We selected the site of aliasing in the renal artery on the color flow image as the region of interest (ROI) (Figs. 1a, b and 2a, b), which was evaluated to be high peak systolic velocity on spectral Doppler. The site and severity of the RAS were also found to be significant (pressure gradient .100 mmHg corresponding to .60% lumen narrowing) on DSA. The ROI was defined based on mutual agreement by two of the authors (J.G., K.M.). In cases of disagreement, a third observer participated in the selection to break the impasse. Background. The entire color box excluding the ROI on the color flow image was considered the color background, which included the iliac artery, iliac vein and renal vein in transplant renal CDUS or the aorta and renal vein in native renal CDUS, depending on both flow velocity in those vessels and color priority settings in each particular case. The color pixel values in the background were counted for each given image. These background color pixels represent the ‘‘color noise’’ from which the CNR is calculated. Calculation of contrast-to-noise ratio The expression for the CNR is given by (Shabana et al. 2009) C jm 2mb j ðsÞ 5 qs ffiffiffiffiffiffiffiffiffi N s2s 1s2b
(1)
2
where ms is the mean color pixel value in the ROI, mb is the mean color pixel value in the background, ss is the standard deviation of the color in the ROI, and sb is the standard deviation of the color background. To calculate the CNR in the green channel, all green pixel values from the ROI are extracted and averaged to find the mean green value of the ROI expressed as Gms. Also, all green pixel values from the background are extracted and averaged to find the mean green value of the background, expressed as Gmb. In addition, the standard deviations of all green pixel values from the ROI and background are expressed as Gss and Gsb, respectively. Then, the CNR in the green channel, G-CNR, is calculated with eqn (1). Subsequently, the CNRs in red and blue channels are calculated with the same equation.
1980
Ultrasound in Medicine and Biology
The CNR in color flow imaging can be estimated from knowledge of the means and variances of the velocity estimates within the stenosis and the background without stenosis, respectively. The numerator in the equation represents the color contrast between the ROI as high-velocity flow at the stenosis and the background as flow at non-stenotic vessels. The larger the value of the CNR, the better is the performance. Statistical analysis All values in each of the red, green and blue channels were expressed as means and standard deviations. Initially, the Kolmogorov-Smirnov test was employed to verify the normal distributions of all data in each color channel. Then a single-factor analysis of variance was applied to determine if any CNR differences existed among all the channels. Given the presence of a difference, unpaired two-tailed t-tests were applied to analyze differences in CNR in each color channel (red, green and blue) between green tagging and standard color mapping. We used a threshold of p , 0.05 to reject the null hypothesis that green tagging did not differ from standard color mapping in displaying aliasing at the site of RAS. RESULTS Our study included a total of 20 patients with hemodynamically significant RAS (arterial lumen reduction .60%) initially detected by renal CDUS and confirmed with DSA. Fast flow velocities represented by aliasing appeared at the site of stenosis. The Kolmogorov-Smirnov test revealed that all the data were normally distributed (p . 0.05). Further, the singlefactor analysis of variance indicated that there was a significant difference in CNR among these groups (p 5 0.000). The CNRs in the red, green and blue channels on green tagging were 0.97 6 0.80, 4.01 6 1.36 and 0.64 6 0.29, respectively. The CNRs in the red, green and blue channels on standard color mapping were 0.35 6 0.44, 1.11 6 0.41 and 0.51 6 0.19, respectively. Further analysis of the individual colors with a t-test revealed that there was no significant difference in the CNR of the red or blue channel between green tagging and standard color mapping (all p . 0.05). However, there was a statistically significant difference in the CNR of the green channel between green tagging and standard color mapping (p 5 0.00019). The CNR of the green channel on the green tagging image was significantly higher than that on the standard color mapping image (Table 1). DISCUSSION We have retrospectively determined that green tagging has a higher CNR than standard color mapping
Volume 39, Number 11, 2013
Table 1. CNR of each color channel on green tagging and standard color mapping Contrast to noise ratio Color channel Red Green Blue
Green tagging
Standard color mapping
p value
0.97 6 0.80 4.01 6 1.36 0.64 6 0.29
0.35 6 0.44 1.11 6 0.41 0.51 6 0.19
0.43762 0.00019 0.56046
in displaying high-velocity flow at a RAS. Although its high sensitivity in displaying the variation in Doppler velocity and backscatter direction for detecting a leaking valve (Switzer et al. 1987) and stenotic artery (Baxter and Polak 1994; Kuyatt et al. 1993) has been reported, to our knowledge, this is the first quantitative comparison of CNR between the use of green tagging and the use of standard color mapping to display aliasing as a function of high-velocity flow for detecting RAS with renal CDUS. Screening for RAS with CDUS has proven to be technically challenging. Among technical errors, a common cause of false-negatives RAS results is underestimation of the flow velocity by missing sampling Doppler shifts at the most stenotic site in the renal artery, which should have the highest flow velocity (Gao et al. 2010). It is well known that estimating the diameter of the renal artery on gray-scale images is very challenging. Not only is the renal artery usually too small to display accurately, but also the poor contrast on gray-scale images makes distinguishing the renal artery from adjacent tissue difficult. Delineation of the stenotic site of the renal artery improved when color flow imaging was developed. Color-coded flow is superimposed on grayscale images to display blood flow in a given anatomic location. The color representing moving red blood cells stands out on the gray-scale background, which makes visualizing the renal artery easier than in gray-scale images. To date, 2-D color flow imaging is still considered an important tool for observing high-velocity flow produced by arterial lumen narrowing. This narrowing is characterized by spectral broadening on spectral Doppler and aliasing in color Doppler produced by Doppler shifts above the Nyquist limit at the site of the RAS (Tamura et al. 1991). Although aliasing on color flow imaging is not a good quantitative indicator of RAS severity, it is still widely used as a qualitative means of identifying the site of fast flow velocity at an arterial narrowing (Chao et al. 1992), which is used to guide spectral Doppler sampling of the highest flow velocity at the RAS. A wide-band pulsed Doppler spectrum results in a larger degree of variance. It has been suggested that
Green tagging vs. standard color mapping in renal artery stenosis d J. GAO et al.
an area of high variance (green color) might provide useful quantitative information regarding the severity of stenotic, regurgitant and shunt lesions (Utsunomiya et al. 1990). Variance expresses the degree to which velocities within a given sample site differ from the mean velocity. Therefore, green tagging is added to display local spreads in velocity; high values are related to steep velocity gradients or turbulence corresponding to the stenosis. In our study, there was no significant difference in CNR in the red or blue channel between the area of stenosis and the background flow on both color mappings (all p . 0.05). However, the CNR in the green channel on the green tagging image of the flow in the stenosis was significantly higher than that in the standard color mapping image (p 5 0.000). The CNR is not rigorously related to the ideal observer performance; however, it is correlated with the visual impressions of the human observer when the lesions are larger than the noise correlations and the noise is roughly constant throughout the entire sonographic image (Insana and Hall 1994). A high CNR means the signal in the ROI is quite different from background and, therefore, easier to differentiate from background. This analysis results in computation of the CNR in different color mappings. The CNR of the color code is an important quantity that is related to the ability to detect high velocity appearing as aliasing on a color flow image. Our results suggest that there is a significant difference in CNR between green tagging and standard color mapping. Green tagging has a higher CNR than standard color mapping in displaying aliasing as a function of local high-velocity flow at the RAS. Hence, it should be easier to detect a green-tagged Doppler signal than an aliased one represented in standard color. The biggest limitation of this study is its retrospective design. Because of this, we only used green tagging in clinical circumstances where it was difficult to identify a specific area of color-flow aliasing before performing spectral analyses. In those specific cases, we felt it appropriate to try to green tag the flow to improve the localization of stenoses based on prior experience with mitral regurgitation and carotid artery stenoses (Baxter and Polak 1994; Switzer et al. 1987). One could therefore presume that we have introduced a bias where green tagging would be more likely to out-perform stenosis localization using color Doppler aliasing in the cases we included in this study. Further, the amount of improvement in contrast we detected between green tagging and the background color would be a function of the color maps employed, and the perception of improvement could easily change as a function of color map. Each pair of our retrospective cases used the same color Doppler parameters; however, we
1981
had no control over the pre- and post-green tagging color maps. As a consequence, the non-green-tagged red and blue distributions might not be identical in the two cases (see the color maps in Figs. 1 and 2). Thus, one could claim that the contrast-to-noise ratios and the relationships we detected among the three channels in the green-tagged and non-green-tagged images might differ for different color maps. It could also be different for different green-tagging variance thresholds, which are defined by the manufacturer. To evaluate these differences and determine possible optimal settings for different color maps and possible variance thresholds for assigning particular pixels green values would require a much larger, prospective study, much more complex than the retrospective study we performed here. However, given that, we can at least say that green tagging is a very good alternative method for identifying stenoses when standard color Doppler aliasing functions poorly in this capacity. Furthermore, these results suggest that green tagging flow could be considered as an excellent color flow mode for assessing and determining the sites of arterial stenoses before spectral Doppler analysis. Additional limitations of this study include the variations in color Doppler settings during each particular renal sonography and inter-observer variation. In addition, all cases included in the study had hemodynamically significant RAS. We did not identify cases with mild or moderate RAS on CDUS that was confirmed with DSA and/or magnetic resonance angiography during this retrospective review. Moreover, the images in this study were recorded from ultrasound scanners made by one manufacturer. Potentially, our observation may differ from the images recorded with ultrasound machines designed by the other manufacturers. We also used only one green tagging map. It is possible that putting more or less weight on variance could have improved the CNR by optimizing the amount of green in the tagged flows. Finally, a color-blind operator may have trouble distinguishing a green-tagged stenosis from one with normal velocities. Because presumably this operator would know about this deficiency ahead of time, he or she could remap the color flow into one where the variance is displayed in blue. Hence the increased CNR would now be in the blue channel, making the stenosis more visible.
CONCLUSIONS Our results indicate that green tagging has a higher CNR than standard color mapping in displaying aliasing as a function of high-velocity flow. Given this high visibility, green tagging may consequently improve the ability to detect the site of RAS to guide spectral Doppler flow velocity analysis.
1982
Ultrasound in Medicine and Biology
REFERENCES American Institute of Ultrasound in Medicine. Practice guidelines for the performance of renal artery duplex sonography. Ultrasound Med 2009;28:120–124. Baxter GM, Polak J. Variance mapping in color flow imaging: What does it measure. Clin Radiol 1994;49:262–265. Borchers J, Klews PM. Digital processing and evaluation of ultrasound images. Ultraschall Med 1993;14:214–219. Chao K, Moises VA, Shandas R, Elkadi T, Sahn DJ, Weitraub R. Influence of coanda effect on color Doppler jet area and color encoding: In vitro studies using color Doppler flow mapping. Circulation 1992; 85:333–341. Gao J, Shin G, Auh YH, Prince MR, Funderburk J, Ng A, Min R. Pitfalls and sources of error of color Doppler sonogaphy in screening for renovascular hypertension. Int J Nephrol Urol 2010;2:212–223. Hedrick WR, Hykes DL. Color Doppler imaging: Imaging parameters and imaging quality. J Diagn Med Sonogr 1992;8:180–186. Insana MF, Hall TJ. Visual detection efficiency in ultrasound imaging: A framework objective assessment of image quality. J Acoust Soc Am 1994;95:2081–2090. Jun WS, Lee KH, Koo KB, Kim DS, Kim JH, Kim SH, Choi BI, Han JK. A straightforward algorithm for quantification of power Doppler signals. Investig Radiol 2002;37:343–348. Kuyatt BL, Reidy CA, Hui KY, Jordan WH. Quantitation of smooth muscle proliferation in cultured aorta: A color image analysis method for the Macintosh. Anal Quant Cytol Histol 1993;15:83–87.
Volume 39, Number 11, 2013 Lee MY, Grant EG. Sonography in renovascular hypertension. J Ultrasound Med 2002;21:431–441. Rubin JM. Questions and answers. AJR Am J Roentgenol 1995;165: 1003–1004. Shabana W, Bude RO, Rubin JM. Comparison between color Doppler twinkling artifact and acoustic shadowing for renal calculus detection: An in vitro study. Ultrasound Med Biol 2009;35:339–350. Song X, Pogue BW, Jiang S, Doyley MM, Dehghani H, Tosteson TD, Paulsen KD. Automated region detection based on the contrast-tonoise ratio in near-infrared tomography. Appl Opt 2004;43: 1053–1062. Spyridopoulos TN, Kaziani K, Balanika AP, Kalokairinou-Motogna M, Bizimi V, Paianidi I, Baltas CS. Ultrasound as a first line screening tool for the detection of renal artery stenosis: A comprehensive review. Med Ultrason 2010;12:228–232. Switzer DF, Yoganathan AP, Nanda NC, Woo YR, Ridgway AJ. Calibration of color Doppler flow mapping during extreme hemodynamic conditions in vitro: A foundation for a reliable quantitative grading system for aortic incompetence. Circulation 1987;75:837–846. Tamura T, Cobbold RSC, Johnson KW. Quantitative study of steady flow using color Doppler ultrasound. Ultrasound Med Biol 1991; 17:595–605. Techavipoo U, Varghese T. Improvements in elastographic contrast tonoise ratio using spatial-angular compounding. Ultrasound Med Biol 2005;31:529–536. Utsunomiya T, Ogawa T, King SW, Sunada E, Moore GW, Henry WL, Gardin JM. Effect of machine parameters on variance display in Doppler color flow mapping. Am Heart J 1990;120:1395–1402.
http://dx.doi.org/10.1016/j.ultrasmedbio.2013.05.006
d
Original Contribution GREEN TAGGING IN DISPLAYING COLOR DOPPLER ALIASING: A COMPARISON TO STANDARD COLOR MAPPING IN RENAL ARTERY STENOSIS JING GAO,* KEVIN MENNITT,* LILY BELFI,* YUAN-YI ZHENG,y ZONG CHEN,z and JONATHAN M. RUBINx * Department of Radiology, New York-Presbyterian Hospital, Weill Cornell Medical College, New York, New York, USA; y Institute of Ultrasound Imaging, The Second Affiliated Hospital, Chongqing Medical University, Chongqing, China; z Department of Computer Science, Fairleigh Dickenson University, Teaneck, New Jersey, USA; and x Department of Radiology, University of Michigan, Ann Arbor, Michigan, USA (Received 12 December 2012; revised 7 May 2013; in final form 14 May 2013)
Abstract—To quantitatively assess the contrast-to-noise ratio (CNR) of green tagging and standard color flow images in displaying fast flow velocity, we retrospectively reviewed 20 cases of hemodynamically significant renal artery stenosis (RAS) detected by renal color Doppler ultrasound and confirmed with digital subtraction angiography. At the site of RAS, blood flow with high velocity that appeared as aliasing on color flow images was computationally analyzed with both green tagging and standard color mapping. To assess the difference in the CNR between normal background flow and the aliased signal as a function of visualizing aliasing between the two color mappings, we used GetColorpixels (Chongqing Medical University, Chongqing, China) to count the values in the color channels after segmenting color pixels from gray-scale pixels. We then calculated the CNR in each color channel—red, green, and blue (RGB)—in the aliasing region on green tagging and standard color mapping. The CNRs in the red, green and blue channels were 0.35 ± 0.44, 1.11 ± 0.41 and 0.51 ± 0.19, respectively, on standard color mapping, and 0.97 ± 0.80, 4.01 ± 1.36 and 0.64 ± 0.29, respectively, on green tagging. We used a singlefactor analysis of variance and two-tailed t-test to assess the difference in CNR in each color channel between the two color mappings at the site of RAS. With these comparisons, there was no significant difference in the CNR in the red or blue channel between green tagging and standard color mapping (p . 0.05). However, there was a statistically significant difference in the CNR in the green channel between the two color mappings (p 5 0.00019). Furthermore, the CNR measured in the green channel on the green tagging image was significantly higher than the CNRs in all other color channels on both color mapping images (p 5 0.000). Hence, we conclude that green tagging has significantly higher visibility as a function of high-velocity flow than standard color mapping. The use of green tagging may improve the ability to detect RAS with color Doppler ultrasound. (E-mail: [email protected]) Ó 2013 World Federation for Ultrasound in Medicine & Biology. Key Words: Color Doppler ultrasound, Contrast-to-noise ratio, Green tagging, Renal artery stenosis.
represents local high velocities that can be used to identify areas of potential stenosis. Thus, given the proper settings for color Doppler frequency and other imaging parameters (total color gain, PRF, wall filter) (Utsunomiya et al. 1990), the identification of local high velocities strongly suggests a reduction of the cross-sectional area of the artery being imaged, that is, arterial luminal narrowing. The velocities increase as a result of renal auto-regulation to maintain constantvolume flow through the narrowed vessel (Rubin 1995). Therefore, the site of aliasing, as a function of highvelocity flow, is suspicious for RAS. Most importantly, the site of aliasing is highly suggestive of an optimal location for placing a spectral Doppler range gate for measuring the local Doppler spectrum.
INTRODUCTION Color Doppler ultrasound (CDUS) is commonly used in screening for renal artery stenosis (RAS), a condition characterized by clinically uncontrolled hypertension and/or renal failure (Lee and Grant 2002; Spyridopoulos et al. 2010). During renal CDUS, color flow imaging is used to assess mean flow velocity in renal vessels. When the pulse repetition frequency (PRF) is appropriately set for blood flow in the kidney and renal artery, local aliasing on color flow imaging Address correspondence to: Yuan-Yi Zheng, Institute of Ultrasound Imaging, The Second Affiliated Hospital, Chongqing Medical University, 76 Lijiang Road, Yuzhong District, Chingqing, China 400010. E-mail: [email protected] 1976
Green tagging vs. standard color mapping in renal artery stenosis d J. GAO et al.
The color flow image is not limited to display of mean velocity. Additional information, such as the spread of velocities within the sampled region, can also be depicted. Therefore, variance color mapping (commonly coded green, hence green tagging) is specifically designed to improve the sensitivity of displaying the variation in Doppler shifts within a sampling region, as a function of Doppler velocity at the region of interest. Thus, in a variance map, the local intensity of green in the 2-D color Doppler image represents a statistical measure of the spread of the velocities around each local mean velocity estimate (Hedrick and Hykes 1992). This is particularly useful in distinguishing areas of turbulent flow, and it was initially used in echocardiography to improve the detection of minimal valve regurgitation jets (Switzer et al. 1987). Green tagging has also been used in vascular sonography to delineate carotid artery stenosis (Baxter and Polak 1994). Because green tagging has been found to be of great value in displaying Doppler shift changes in cardiovascular ultrasound, we hypothesized that the contrast-tonoise ratio (CNR) of green tagging would be better than that of standard color Doppler mapping in detecting fast flow velocity at the site of RAS. The ultimate goal of this comparative study was to evaluate which color mapping has a greater ability to detect RAS with CDUS. METHODS This retrospective study was approved by the institutional review board (IRB 1201012129) of Weill Cornell Medical College and conducted at the New YorkPresbyterian Hospital, New York. The requirement for informed consent was waived. The study was compliant with the Health Insurance Portability and Accountability Act of 1996. Patients From January 2010 to June 2011, 20 patients with RAS (12 men and 8 women, mean age 5 59 6 14 y, range: 25–80 y) were identified by renal CDUS, using both green tagging and standard color mapping, and confirmed with digital subtraction angiography (DSA). All 20 cases with hemodynamically significant RAS (.60% arterial lumen narrowing, 10 with native RAS and 10 with transplant RAS) had a medical history of uncontrolled hypertension (rest blood pressure .150/90 mm Hg, n 5 11), renal failure (serum creatinine .1.1 mg/dL, n 5 4) or both (n 5 5). We reviewed static images and cine loops of renal CDUS recordings and static images of the DSA through a web-based Picture Archiving and Communication System (PACS, Centricity Enterprise Web V2.1, GE Medical Systems, Milwaukee, WI, USA). We searched our institutional electronic medical records system for
1977
pertinent subject demographic information including sex, age and reason for undergoing renal CDUS and DSA. We included 20 cases with RAS detected with both standard color mapping and green tagging with the same color Doppler parameters. All cases of RAS were confirmed with DSA. Renal color Doppler sonography No specific preparation for renal transplant CDUS was required before examinations. All patients who underwent native renal CDUS were requested to fast for 6–8 h before their ultrasound examinations. Patients were placed in the supine position for renal transplant CDUS and left or right decubitus position for native renal CDUS. A curved linear array transducer (4C1) or sector transducer (4V1) with multi-frequencies of 2–4 MHz (Acuson Sequoia 512 or S2000, Siemens Medical Solutions, Mountain View, CA, USA) was used to acquire color Doppler images. All scans were performed according to the American Institute of Ultrasound in Medicine (2009) practice guideline for the performance of renal artery duplex sonography. We started renal sonography by measuring renal size, observing cortical echogenicity and evaluating renal hydronephrosis, calculus or mass on gray-scale imaging. Color flow imaging was then performed to assess renal vasculature. Color Doppler settings including PRF, wall filter and overall color gain were adjusted from minimum to maximum to observe arteries in the renal parenchyma with slow flow and to visualize renovascular abnormalities with fast flow, that is, intra-renal RAS. Detection of vascular abnormalities in the kidney is highly dependent on the color Doppler settings. A low PRF may cause Doppler aliasing that overlaps high-velocity flow, resulting in a decrease in the distinction between aliased fast flow velocity at the stenotic artery and slow velocity flow in a normal artery. On the other hand, a high PRF could suppress slow flow in intra-renal vessels to make fast flow velocities at the stenosis obvious (Gao et al. 2010). In our institution, color Doppler was pre-set with standard color mapping in all ultrasound scanners. Green tagging was manually selected for recognizing fast flow velocities when the aliasing was not recognizable on standard color mapping images in the cases highly suspicious for RAS. It is known that machine parameters strongly affect the display of fast flow velocities as aliasing in color flow imaging (Utsunomiya et al. 1990). To minimize variations in color values produced by different color Doppler parameter settings and ultrasound machine design, we only included 20 cases of RAS imaging acquired using both standard color mapping (Fig. 1a) and green tagging (Fig. 2a) with the same color Doppler parameters on the same ultrasound scanner.
1978
Ultrasound in Medicine and Biology
Volume 39, Number 11, 2013
Fig. 1. Renal color Doppler sonography was performed on a renal transplant patient who developed uncontrolled hypertension. The arterial anastomosis, transplanted renal artery and iliac artery were displayed on a zoomed color Doppler image with standard color mapping (a). The rectangular box is the region of interest (ROI), in which questionable color Doppler aliasing appears at the site of a suspected arterial stenosis (a). Also shown is the zoomed ROI with color segmentation (b) processed with software (GetColorpixels). To calculate the contrast-to-noise ratio of the standard color mapping image, color pixels in the ROI with suspicious high velocity and the background (entire color box excluding ROI) are counted.
A spectral Doppler gate of 2–4 mm was placed at the center of the artery to measure the maximum velocity with Doppler angle correction by adjusting the emission sound beam to be as parallel as possible to the flow direction (at least ,60 ). The spectral Doppler waveforms were recorded for flow velocity measurements. We also assessed the abdominal aorta (in native renal sonography) and the iliac artery (in renal transplant sonography) with color flow imaging and spectral Doppler. Peak systolic velocity (PSV) was manually measured at the renal artery (proximal, mid and distal portions), abdominal aorta and iliac artery with electronic calipers. Special attention was paid to an artery with focal aliasing throughout the scan. The ratio of the PSV of the stenotic artery to that of the
artery proximal to the stenosis was calculated. A diagnosis of RAS was considered when the renal artery/aorta (iliac artery in renal transplant) PSV ratio was greater than 3.5 and the PSV at the renal artery was .1.8 m/s in native kidney (.2.6 m/s in transplanted kidney) (Gao et al. 2010). All 20 cases with RAS were confirmed with DSA. Quantification of color Doppler signal To count pixels in each color channel, color pixels must be separated from gray-scale pixels (Borchers and Klews 1993). For standard color mapping, the colorcoded image representing blood flow direction and mean velocity is overlaid on the gray-scale (B-mode)
Fig. 2. With the same color Doppler priority settings as used for standard color mapping in Figure 1, green tagging (green arrow) was applied to color flow images at the same anatomic location as in Figure 1. (Note that the red/blue color distributions slightly differ between Figs. 1 and 2.) It is clearly noted that significant aliasing appears as green at the proximal transplanted renal artery. The rectangular box contains the region of interest (ROI) (a). A high flow velocity appears in the ROI with color segmentation (b). The background for calculating the contrast-to-noise ratio is the entire color box excluding the ROI.
Green tagging vs. standard color mapping in renal artery stenosis d J. GAO et al.
image to form an output digital color Doppler image display on the screen of an ultrasound scanner. Moreover, fast flow with a wide range of flow velocities was tagged as green that is overlaid on a standard color mapping image to form a green tagging image. Hence, green tagging is considered a velocity tagging function that emphasizes a selected range of velocities with a high variance by highlighting regions of flow containing such velocities in green. On the ultrasound scanner, the color bar adjacent to the displayed image indicates in green (Fig. 2a, green arrow) that portion of the velocity scale selected for tagging. For offline image processing, de-identified (anonymized) static images of green tagging and standard color mapping with TIFF format were transferred to a personal computer. Ultrasound image analysis software (GetColorpixels, Chongqing Medical University, Chongqing, China) was developed to quantify the color pixels. Color pixels representing color Doppler signals need to be separated from the background gray pixels in color Doppler images. To avoid mis-registration during this segmentation, a threshold value was required to adequately distinguish between gray and color pixels (Jun et al. 2002). In a standard Doppler image, the background is mapped with pixels with little variance in R, G, B components, whereas the Doppler signals were mapped with pixels with large differences in R, G, B components, which could be used to differentiate the different pixels by setting a threshold. If the value (maximum value of R/G/B – minimum value of R/G/B) is smaller than this threshold, the pixel is considered a background pixel; otherwise, it is considered a Doppler signal pixel. We set a threshold of 50, which completely deleted the gray pixels (Figs. 1b and 2b), allowing us to segment the color pixels (Jun et al. 2002). After segmenting color pixels, we used the Image J (National Institutes of Health, Bethesda, MD, USA) histogram to calculate the pixels in the region that was previously gray. In all cases, a zero value was measured, which indicated that all gray-scale pixels had been completely deleted, leaving only color pixels in each image. It was now possible to count the segmented color pixels. This threshold may delete some color pixels during the processing; however, as long as we used the same threshold setting for all cases, there should be no effect on the comparison of color pixel values in each color channel between standard color mapping and green tagging. Finally, the color values in each color channel, R, G and B, was measured with the software GetColorpixels at the region of interest and background to calculate the CNR. Contrast-to-noise ratio of green tagging and standard color mapping The CNR is a standard method for estimating the ability to detect targets in diagnostic imaging (Song
1979
et al. 2004; Techavipoo and Varghese 2005). In screening for RAS, color flow imaging is used to detect the location of aliasing as a function of high-velocity flow at the site of arterial stenosis, which is further evaluated with spectral Doppler velocity analysis. Therefore, it is important to recognize local aliasing on color Doppler images. Region of interest We selected the site of aliasing in the renal artery on the color flow image as the region of interest (ROI) (Figs. 1a, b and 2a, b), which was evaluated to be high peak systolic velocity on spectral Doppler. The site and severity of the RAS were also found to be significant (pressure gradient .100 mmHg corresponding to .60% lumen narrowing) on DSA. The ROI was defined based on mutual agreement by two of the authors (J.G., K.M.). In cases of disagreement, a third observer participated in the selection to break the impasse. Background. The entire color box excluding the ROI on the color flow image was considered the color background, which included the iliac artery, iliac vein and renal vein in transplant renal CDUS or the aorta and renal vein in native renal CDUS, depending on both flow velocity in those vessels and color priority settings in each particular case. The color pixel values in the background were counted for each given image. These background color pixels represent the ‘‘color noise’’ from which the CNR is calculated. Calculation of contrast-to-noise ratio The expression for the CNR is given by (Shabana et al. 2009) C jm 2mb j ðsÞ 5 qs ffiffiffiffiffiffiffiffiffi N s2s 1s2b
(1)
2
where ms is the mean color pixel value in the ROI, mb is the mean color pixel value in the background, ss is the standard deviation of the color in the ROI, and sb is the standard deviation of the color background. To calculate the CNR in the green channel, all green pixel values from the ROI are extracted and averaged to find the mean green value of the ROI expressed as Gms. Also, all green pixel values from the background are extracted and averaged to find the mean green value of the background, expressed as Gmb. In addition, the standard deviations of all green pixel values from the ROI and background are expressed as Gss and Gsb, respectively. Then, the CNR in the green channel, G-CNR, is calculated with eqn (1). Subsequently, the CNRs in red and blue channels are calculated with the same equation.
1980
Ultrasound in Medicine and Biology
The CNR in color flow imaging can be estimated from knowledge of the means and variances of the velocity estimates within the stenosis and the background without stenosis, respectively. The numerator in the equation represents the color contrast between the ROI as high-velocity flow at the stenosis and the background as flow at non-stenotic vessels. The larger the value of the CNR, the better is the performance. Statistical analysis All values in each of the red, green and blue channels were expressed as means and standard deviations. Initially, the Kolmogorov-Smirnov test was employed to verify the normal distributions of all data in each color channel. Then a single-factor analysis of variance was applied to determine if any CNR differences existed among all the channels. Given the presence of a difference, unpaired two-tailed t-tests were applied to analyze differences in CNR in each color channel (red, green and blue) between green tagging and standard color mapping. We used a threshold of p , 0.05 to reject the null hypothesis that green tagging did not differ from standard color mapping in displaying aliasing at the site of RAS. RESULTS Our study included a total of 20 patients with hemodynamically significant RAS (arterial lumen reduction .60%) initially detected by renal CDUS and confirmed with DSA. Fast flow velocities represented by aliasing appeared at the site of stenosis. The Kolmogorov-Smirnov test revealed that all the data were normally distributed (p . 0.05). Further, the singlefactor analysis of variance indicated that there was a significant difference in CNR among these groups (p 5 0.000). The CNRs in the red, green and blue channels on green tagging were 0.97 6 0.80, 4.01 6 1.36 and 0.64 6 0.29, respectively. The CNRs in the red, green and blue channels on standard color mapping were 0.35 6 0.44, 1.11 6 0.41 and 0.51 6 0.19, respectively. Further analysis of the individual colors with a t-test revealed that there was no significant difference in the CNR of the red or blue channel between green tagging and standard color mapping (all p . 0.05). However, there was a statistically significant difference in the CNR of the green channel between green tagging and standard color mapping (p 5 0.00019). The CNR of the green channel on the green tagging image was significantly higher than that on the standard color mapping image (Table 1). DISCUSSION We have retrospectively determined that green tagging has a higher CNR than standard color mapping
Volume 39, Number 11, 2013
Table 1. CNR of each color channel on green tagging and standard color mapping Contrast to noise ratio Color channel Red Green Blue
Green tagging
Standard color mapping
p value
0.97 6 0.80 4.01 6 1.36 0.64 6 0.29
0.35 6 0.44 1.11 6 0.41 0.51 6 0.19
0.43762 0.00019 0.56046
in displaying high-velocity flow at a RAS. Although its high sensitivity in displaying the variation in Doppler velocity and backscatter direction for detecting a leaking valve (Switzer et al. 1987) and stenotic artery (Baxter and Polak 1994; Kuyatt et al. 1993) has been reported, to our knowledge, this is the first quantitative comparison of CNR between the use of green tagging and the use of standard color mapping to display aliasing as a function of high-velocity flow for detecting RAS with renal CDUS. Screening for RAS with CDUS has proven to be technically challenging. Among technical errors, a common cause of false-negatives RAS results is underestimation of the flow velocity by missing sampling Doppler shifts at the most stenotic site in the renal artery, which should have the highest flow velocity (Gao et al. 2010). It is well known that estimating the diameter of the renal artery on gray-scale images is very challenging. Not only is the renal artery usually too small to display accurately, but also the poor contrast on gray-scale images makes distinguishing the renal artery from adjacent tissue difficult. Delineation of the stenotic site of the renal artery improved when color flow imaging was developed. Color-coded flow is superimposed on grayscale images to display blood flow in a given anatomic location. The color representing moving red blood cells stands out on the gray-scale background, which makes visualizing the renal artery easier than in gray-scale images. To date, 2-D color flow imaging is still considered an important tool for observing high-velocity flow produced by arterial lumen narrowing. This narrowing is characterized by spectral broadening on spectral Doppler and aliasing in color Doppler produced by Doppler shifts above the Nyquist limit at the site of the RAS (Tamura et al. 1991). Although aliasing on color flow imaging is not a good quantitative indicator of RAS severity, it is still widely used as a qualitative means of identifying the site of fast flow velocity at an arterial narrowing (Chao et al. 1992), which is used to guide spectral Doppler sampling of the highest flow velocity at the RAS. A wide-band pulsed Doppler spectrum results in a larger degree of variance. It has been suggested that
Green tagging vs. standard color mapping in renal artery stenosis d J. GAO et al.
an area of high variance (green color) might provide useful quantitative information regarding the severity of stenotic, regurgitant and shunt lesions (Utsunomiya et al. 1990). Variance expresses the degree to which velocities within a given sample site differ from the mean velocity. Therefore, green tagging is added to display local spreads in velocity; high values are related to steep velocity gradients or turbulence corresponding to the stenosis. In our study, there was no significant difference in CNR in the red or blue channel between the area of stenosis and the background flow on both color mappings (all p . 0.05). However, the CNR in the green channel on the green tagging image of the flow in the stenosis was significantly higher than that in the standard color mapping image (p 5 0.000). The CNR is not rigorously related to the ideal observer performance; however, it is correlated with the visual impressions of the human observer when the lesions are larger than the noise correlations and the noise is roughly constant throughout the entire sonographic image (Insana and Hall 1994). A high CNR means the signal in the ROI is quite different from background and, therefore, easier to differentiate from background. This analysis results in computation of the CNR in different color mappings. The CNR of the color code is an important quantity that is related to the ability to detect high velocity appearing as aliasing on a color flow image. Our results suggest that there is a significant difference in CNR between green tagging and standard color mapping. Green tagging has a higher CNR than standard color mapping in displaying aliasing as a function of local high-velocity flow at the RAS. Hence, it should be easier to detect a green-tagged Doppler signal than an aliased one represented in standard color. The biggest limitation of this study is its retrospective design. Because of this, we only used green tagging in clinical circumstances where it was difficult to identify a specific area of color-flow aliasing before performing spectral analyses. In those specific cases, we felt it appropriate to try to green tag the flow to improve the localization of stenoses based on prior experience with mitral regurgitation and carotid artery stenoses (Baxter and Polak 1994; Switzer et al. 1987). One could therefore presume that we have introduced a bias where green tagging would be more likely to out-perform stenosis localization using color Doppler aliasing in the cases we included in this study. Further, the amount of improvement in contrast we detected between green tagging and the background color would be a function of the color maps employed, and the perception of improvement could easily change as a function of color map. Each pair of our retrospective cases used the same color Doppler parameters; however, we
1981
had no control over the pre- and post-green tagging color maps. As a consequence, the non-green-tagged red and blue distributions might not be identical in the two cases (see the color maps in Figs. 1 and 2). Thus, one could claim that the contrast-to-noise ratios and the relationships we detected among the three channels in the green-tagged and non-green-tagged images might differ for different color maps. It could also be different for different green-tagging variance thresholds, which are defined by the manufacturer. To evaluate these differences and determine possible optimal settings for different color maps and possible variance thresholds for assigning particular pixels green values would require a much larger, prospective study, much more complex than the retrospective study we performed here. However, given that, we can at least say that green tagging is a very good alternative method for identifying stenoses when standard color Doppler aliasing functions poorly in this capacity. Furthermore, these results suggest that green tagging flow could be considered as an excellent color flow mode for assessing and determining the sites of arterial stenoses before spectral Doppler analysis. Additional limitations of this study include the variations in color Doppler settings during each particular renal sonography and inter-observer variation. In addition, all cases included in the study had hemodynamically significant RAS. We did not identify cases with mild or moderate RAS on CDUS that was confirmed with DSA and/or magnetic resonance angiography during this retrospective review. Moreover, the images in this study were recorded from ultrasound scanners made by one manufacturer. Potentially, our observation may differ from the images recorded with ultrasound machines designed by the other manufacturers. We also used only one green tagging map. It is possible that putting more or less weight on variance could have improved the CNR by optimizing the amount of green in the tagged flows. Finally, a color-blind operator may have trouble distinguishing a green-tagged stenosis from one with normal velocities. Because presumably this operator would know about this deficiency ahead of time, he or she could remap the color flow into one where the variance is displayed in blue. Hence the increased CNR would now be in the blue channel, making the stenosis more visible.
CONCLUSIONS Our results indicate that green tagging has a higher CNR than standard color mapping in displaying aliasing as a function of high-velocity flow. Given this high visibility, green tagging may consequently improve the ability to detect the site of RAS to guide spectral Doppler flow velocity analysis.
1982
Ultrasound in Medicine and Biology
REFERENCES American Institute of Ultrasound in Medicine. Practice guidelines for the performance of renal artery duplex sonography. Ultrasound Med 2009;28:120–124. Baxter GM, Polak J. Variance mapping in color flow imaging: What does it measure. Clin Radiol 1994;49:262–265. Borchers J, Klews PM. Digital processing and evaluation of ultrasound images. Ultraschall Med 1993;14:214–219. Chao K, Moises VA, Shandas R, Elkadi T, Sahn DJ, Weitraub R. Influence of coanda effect on color Doppler jet area and color encoding: In vitro studies using color Doppler flow mapping. Circulation 1992; 85:333–341. Gao J, Shin G, Auh YH, Prince MR, Funderburk J, Ng A, Min R. Pitfalls and sources of error of color Doppler sonogaphy in screening for renovascular hypertension. Int J Nephrol Urol 2010;2:212–223. Hedrick WR, Hykes DL. Color Doppler imaging: Imaging parameters and imaging quality. J Diagn Med Sonogr 1992;8:180–186. Insana MF, Hall TJ. Visual detection efficiency in ultrasound imaging: A framework objective assessment of image quality. J Acoust Soc Am 1994;95:2081–2090. Jun WS, Lee KH, Koo KB, Kim DS, Kim JH, Kim SH, Choi BI, Han JK. A straightforward algorithm for quantification of power Doppler signals. Investig Radiol 2002;37:343–348. Kuyatt BL, Reidy CA, Hui KY, Jordan WH. Quantitation of smooth muscle proliferation in cultured aorta: A color image analysis method for the Macintosh. Anal Quant Cytol Histol 1993;15:83–87.
Volume 39, Number 11, 2013 Lee MY, Grant EG. Sonography in renovascular hypertension. J Ultrasound Med 2002;21:431–441. Rubin JM. Questions and answers. AJR Am J Roentgenol 1995;165: 1003–1004. Shabana W, Bude RO, Rubin JM. Comparison between color Doppler twinkling artifact and acoustic shadowing for renal calculus detection: An in vitro study. Ultrasound Med Biol 2009;35:339–350. Song X, Pogue BW, Jiang S, Doyley MM, Dehghani H, Tosteson TD, Paulsen KD. Automated region detection based on the contrast-tonoise ratio in near-infrared tomography. Appl Opt 2004;43: 1053–1062. Spyridopoulos TN, Kaziani K, Balanika AP, Kalokairinou-Motogna M, Bizimi V, Paianidi I, Baltas CS. Ultrasound as a first line screening tool for the detection of renal artery stenosis: A comprehensive review. Med Ultrason 2010;12:228–232. Switzer DF, Yoganathan AP, Nanda NC, Woo YR, Ridgway AJ. Calibration of color Doppler flow mapping during extreme hemodynamic conditions in vitro: A foundation for a reliable quantitative grading system for aortic incompetence. Circulation 1987;75:837–846. Tamura T, Cobbold RSC, Johnson KW. Quantitative study of steady flow using color Doppler ultrasound. Ultrasound Med Biol 1991; 17:595–605. Techavipoo U, Varghese T. Improvements in elastographic contrast tonoise ratio using spatial-angular compounding. Ultrasound Med Biol 2005;31:529–536. Utsunomiya T, Ogawa T, King SW, Sunada E, Moore GW, Henry WL, Gardin JM. Effect of machine parameters on variance display in Doppler color flow mapping. Am Heart J 1990;120:1395–1402.