Author + information
- Received August 10, 2015
- Revision received October 8, 2015
- Accepted October 15, 2015
- Published online September 1, 2016.
- Eun-Ah Park, MD, PhDa,
- Whal Lee, MD, PhDa,∗ (, )
- Sang Joon Park, PhDa,
- Yeo Koon Kim, MDa and
- Ho Young Hwang, MD, PhDb
- aDepartment of Radiology, Seoul National University Hospital, Seoul, Republic of Korea
- bDepartment of Thoracic Surgery, Seoul National University Hospital, Seoul, Republic of Korea
- ↵∗Reprint requests and correspondence:
Dr. Whal Lee, Department of Radiology, Seoul National University Hospital, 28 Yongon-Dong, Chongno-Gu, Seoul 110-744, Republic of Korea.
Objectives The goal of this study was to assess the effect of coronary artery diameter on luminal attenuation and the correlation between the transluminal attenuation gradient (TAG) and transluminal diameter gradient (TDG) on computed tomography (CT) coronary angiography.
Background Recent studies have reported promising results of TAG in detecting significant stenosis. However, because of the intrinsic nature of CT reconstruction algorithms, luminal attenuation may be affected by vessel diameter.
Methods In this 3-part study, phantom simulating vessels of various diameters immersed in different contrast mixtures were scanned, and intraluminal attenuations were measured. In addition, dynamic volume CT scanning was performed in 3 mongrel dogs (untreated, a stenosis model, and an occlusion model) using 320-row area detector computed tomography and intraluminal attenuations, and TAGs were calculated at each temporal scan and compared. In a separate clinical study, TAGs and TDGs of 152 coronary arteries from 62 patients who underwent 320-row area detector computed tomography coronary angiography and invasive angiography were measured and compared.
Results Intraluminal attenuation of phantom vessels gradually decreased along with a decrease in diameter. Animal studies revealed that the peak attenuation of distal smaller coronary arteries did not reach that of proximal larger coronary arteries: 55.2% to 78.1% peak attenuation of proximal coronary arteries. No differences in TAG were found between stenotic and normal left circumflex arteries at temporal scans (all, p > 0.05). The clinical study demonstrated significant correlation between TAG and TDG (r = 0.580; p < 0.0001).
Conclusions Intraluminal attenuation was shown to decrease with diminution of vessel diameters. In addition, TAG exhibited a significant correlation with TDG, implying that TAG may be a secondary result because of differences in diameters.
According to a recent statistical update report from the American Heart Association, despite a significant reduction in the death rate from cardiovascular diseases, the leading cause of morbidity and mortality in 2020 will still be cardiovascular disease (1). At present, coronary computed tomography angiography (CTA) is widely used for the assessment of patients with suspected coronary artery disease (CAD); this test is a highly accurate noninvasive technique for the diagnosis of CAD, capable of providing detailed information of coronary artery anatomy, plaque characteristics, and the degree of stenosis (2). However, coronary CTA has an important limitation in that it cannot determine functionally significant stenosis (3). Thus, there have been several attempts to add functional information to anatomic CAD assessment so as to determine the hemodynamic significance of CAD lesions, including computed tomography (CT) perfusion (4), vulnerable plaque characterization (5,6), the intracoronary transluminal attenuation gradient (TAG) (7–9), corrected coronary opacification (10), and CT fractional flow reserve (FFR) (11).
TAG, initially introduced by Steigner et al. (7), has demonstrated the ability to differentiate hemodynamically significant CAD in several previous clinical studies (8,12). TAG is based on the concept that if all the factors which affect coronary artery attenuation were kept constant, including left ventricular function, contrast media volume and concentration, and contrast bolus rates, the gradual diminution gradient of intraluminal attenuation from the proximal to distal segment could then be used as a surrogate to resting coronary flow (13). However, intraluminal coronary attenuation has been shown to be affected by the diameter of the coronary artery lumen because attenuation of smaller arteries may decrease because of the intrinsic point-spread function of CT scanning using reconstruction algorithms (14). Thus, we hypothesized that intraluminal attenuation may decrease as the diameter decreases, even though it is filled with consistently concentrated contrast media, and that TAG may be related to the transluminal diameter gradient (TDG).
The present study therefore evaluated the effect of coronary artery diameter on luminal attenuation by using a vessel phantom, an animal study with a canine coronary artery stenosis model, and a clinical study to investigate the correlation between TAG and TDG.
A vessel phantom was built with 25 tubular holes in a 5-cm thick polyethylene disc (Online Figure 1). The diameter of each hole was 5.0, 4.5, 4.0, 3.5, 3.0, and 2.9 mm to 1.0 mm with 0.1-mm intervals, simulating vessels of different diameters. The phantom was immersed in 2 different mixtures of an iodine contrast agent (Ultravist, Shering, Berlin, Germany) and saline, approximating attenuations of 800 and 600 HU at 100 kVp, with all tubular holes completely filled with the contrast mixture. CT scans were performed with the phantom immersed in the 600-HU contrast mixture using 80, 100, and 120 kVp. Thereafter, the phantom was washed and immersed in the 800-HU contrast mixture and scanned using 100 kVp. CT scans were performed 5 times in each protocol. All examinations were performed with a 320-row area detector computed tomography (320-ADCT) scanner (Aquilion ONE, Toshiba Medical Systems, Tokyo, Japan) with a field of view of 120 × 120 mm and a matrix size of 512 × 512 mm, resulting in a pixel size of 0.23 mm. The intraluminal attenuation of each hole was measured using dedicated software for coronary CTA (Xelis Cardiac, INFINITT Healthcare, Seoul, Republic of Korea) by 1 board-certified radiologist with 3 years of experience in cardiovascular imaging (Y.K.K.).
The centerline of each tubular hole was semi-automatically selected, and the attenuation was measured at every pixel through the centerline. Intraluminal attenuation was calculated as the average attenuation of 9 pixels (3 × 3 pixels) around the centerline pixel indicating a square-shaped region of interest (ROI) of 0.70 mm in length and 0.49 mm2 in area.
The animal experiment, which established 1 stenosis model, 1 occlusion model, and an untreated mongrel as a control subject, was approved by our Institutional Animal Care and Use Committee. Two mongrel dogs who weighed 30 to 35 kg were anesthetized with a subcutaneous injection of a 15-mg/kg mixture of zolazepam (Zoletil, Yuhan Corp., Seoul, Republic of Korea), 5 to 10 mg/kg of xylazine hydrochloride (Rompun, Bayer Korea, Seoul, Republic of Korea), and 0.02 to 0.04 mg/kg of atropine sulfate (Atropine, DAHIHAN Pharm, Co., Ltd., Seoul, Republic of Korea); they were intubated and mechanically ventilated during preparation. After vascular cut-downs, 5-F sheaths were placed in the left carotid artery. Left thoracotomy was performed in the fifth intercostal space, and the pericardium was excised. The left circumflex coronary artery was carefully dissected and ligated completely for the occlusion model; it was ligated with a 5-F micro-introducer with subsequent removal of the catheter for the stenosis model. Through the sheath in the carotid artery, catheter angiography of the left coronary artery was performed to confirm the stenosis and occlusion of the coronary artery (Online Figure 2). After confirmation of stenosis and occlusion, the thoracotomy incision was closed, and the carotid artery was repaired.
Dynamic volume CT imaging protocol
Dynamic volume CT examinations were performed 15 times in total in the 3 mongrel dogs using 320-ADCT (Toshiba Medical Systems) under anesthesia 7 days after modeling: 4 times in the dog with no treatment, 6 times in the stenosis model, and 5 times in the occlusion model. Intravenous esmolol (1 to 2 mg) was injected to achieve a heart rate of <80 beats/min. Mean heart rate was maintained from 70 to 80 beats/min during the scans; 20 ml of a nonionic contrast medium (Iomeron 400, Bracco Diagnostics, Milan, Italy) was injected into the right cephalic vein at 4 ml/s, followed by 20 ml of normal saline at the same flow rate. CT scanning was initiated 10 s after contrast administration and performed every heartbeat during the 40 s by using the prospective ECG triggering technique. CT parameters were as follows: tube voltage, 100 kVp; field of view, 122 × 122 mm; matrix size, 512 × 512 mm; and pixel size, 0.24 mm. SureExposure (Toshiba Medical Systems) with a target SD of 25 was used. Reconstructed phases were selected manually at each temporal scan to minimize motion artifacts.
All measurements were performed by a board-certified cardiovascular radiologist with 10 years of experience (P.E.A.) using 3-dimensional dedicated software (Xelis Cardiac, INFINITT Healthcare). Mean vessel attenuation at 7 different anatomic regions at each temporal scan was measured using a 2 cm2-circular ROI for the aorta and a 1 mm2circular ROI for the coronary arteries: proximal ascending aorta, left anterior descending coronary artery (LAD) proximal (1.0 cm from the ostium), LAD mid (3.0 cm from the ostium), LAD distal (5.5 mm from the ostium), left circumflex coronary artery (LCx) proximal (1.5 cm from the ostium), LCx distal (7.0 cm from the ostium), and posterior descending artery proximal (9.0 cm from the ostium). Thereafter, the percentage of the peak attenuation of each region was calculated and compared with the peak attenuation of the ascending aorta. Vessel diameter at each coronary region analyzed was also calculated. In addition, TAG was manually measured by using a method described in a previous study (8). The method included 5 heartbeat scans previous to that at peak aortic enhancement and 8 images after peak aortic enhancement, to minimize the effect of different hemodynamic statuses. The total length of the coronary artery measured for TAG was 5.0 cm for LAD and 8.5 cm for LCx.
Our institutional review board approved the clinical study protocol of this retrospective study and waived the requirement for informed consent.
From January 2012 to July 2012, patients who underwent coronary CTA and invasive coronary angiography within 90 days were included. Exclusion criteria were as follows: 1) inadequate CT image quality due to motion artifacts or beam-hardening artifacts by a coronary stent; 2) adaptive multisegment reconstruction due to a fast heart rate; 3) diffuse severe coronary calcification, which made measurement of the exact intraluminal attenuation difficult; 4) hypoplastic coronary arteries with proximal coronary artery diameters <3 mm; 5) left main coronary artery stenosis >50%; and 6) patients with acute coronary syndrome within 90 days. Ultimately, 152 coronary arteries of 62 patients (32 men; mean age 65.9 years; age range 27 to 87 years) were included for analysis.
All CT examinations were acquired using a 320-ADCT system (Aquilion ONE, Toshiba Medical Systems) with mid-diastolic prospective scanning, as described in detail in a previous study (15). The pixel size was 0.33 mm.
Two board-certified cardiovascular radiologists (W.L. [with 11 years of experience] and Y.K.K. [with 3 years of experience]) reviewed and analyzed the coronary CTA images in consensus. Intraluminal attenuation was measured using dedicated software for coronary CTAs (Xelis Cardiac, INFINITT Healthcare). The centerline of the coronary artery was semi-automatically selected, and if needed, the centerline was edited by manual drawing on a perpendicular plane to the coronary artery at each level (Figure 1). Intraluminal attenuation at the centerline (Hounsfield units), luminal diameter (millimeters), and luminal area (square millimeters) were automatically measured at each pixel through the centerlines of the coronary arteries. Measurements were performed until the luminal diameter of the distal coronary artery reached 1.5 mm. Intraluminal attenuation was calculated in the same manner as described in the phantom study.
The definition of TAG was applied from a previous study (8). TAG was determined from the change of attenuation (Hounsfield units) per 10 mm of length of the coronary artery and defined as the linear regression coefficient between the attenuation (Hounsfield units) and length from the ostium (millimeters). A TAG ≥20 was considered to indicate a TAG-positive artery.
Similarly, TDG was determined from the change of diameter per 10 mm of length of the coronary artery and defined as the linear regression coefficient between the diameter and length from the ostium. Attenuation and diameter data at the level of stenosis and calcifications were excluded when calculating TAG and TDG (Figure 1).
Coronary angiography analysis
Quantitative analysis of invasive coronary angiography was performed with quantitative coronary angiography (CAAS 5.7 QCA system, Pie Medical Imaging, Maastricht, the Netherlands) by a single experienced observer who was blinded to coronary CTA findings. Diameter stenosis was used as the reference of standard. A diameter stenosis ≥50% was defined as significant stenosis.
All data processing and analyses were performed using SPSS version 22.0 (IBM SPSS Statistics, IBM Corporation, Armonk, New York). Two-tailed p values <0.05 were considered to indicate a significant difference. All statistical analyses were performed nonparametrically because neither clinical nor animal data showed a normal distribution. Data are presented as medians and interquartile ranges. In the animal study, differences in TAG values at each temporal scan were evaluated by using the Kruskal-Wallis test and Dunn’s multiple comparison post hoc test. In the clinical study, the correlation between TAG and TDG was performed by using Spearman correlation analysis. In addition, the Mann-Whitney U test was performed for the comparison of TDG between TAG-positive and TAG-negative arteries.
Phantom study: Intraluminal attenuation according to the diameter of holes
Even though all holes were filled with the same contrast mixture, intraluminal attenuation was observed to gradually decline along with the decreased diameter of the holes (Online Figure 1); the reduction of intraluminal attenuation was much greater as the diameters became smaller. The same findings were observed regardless of the tube voltage used or the concentration of contrast mixture filled in the phantom. Detailed attenuation values are provided in Online Table 1.
Animal study: Change in attenuation and TAG of each coronary model over time
Peak attenuation of distal small coronary arteries was slightly delayed in cases in which LCx was longer than LAD but did not reach that of proximal large coronary arteries. They exhibited only 55.2% to 78.1% of the maximum enhancement of proximal segments, regardless of type of coronary arteries (Table 1, Online Figure 3).
Figure 2 displays TAG values of each coronary model over time. There were significant differences in TAG values starting from the third repeated scan after the attenuation of the ascending aorta reached a peak to the eighth repeated scan (Table 2): between TAG values of normal LAD and occluded LCx in the fifth to eighth repeated scans, and between TAG values of stenotic LCx and occluded LCx in the third to eighth repeated scans (all, p < 0.05) (Figure 3). However, there were no significant differences in TAG values between stenotic and normal LCx or between stenotic LCx and normal LAD (all, p > 0.05).
Clinical study: Association between TAG and TDG
Although the correlation coefficient was modest, a significant correlation was found between TAG and TDG in total arteries (r = 0.580; p < 0.0001) (Figure 4). Separate analysis of significant and nonsignificant stenosis groups also showed a significant correlation between TAG and TDG (r = 0.610 [p < 0.0001] for nonsignificant stenosis; r = 0.565 [p = 0.0001] for significant stenosis). In addition, TAG-positive arteries exhibited significantly greater TDG values in both groups with significant stenosis (p < 0.0001) and without significant stenosis (p < 0.0001) (Table 3). There were no significant differences in TAG values between groups with significant stenosis and without significant stenosis (p = 0.884), whereas a significant difference was found in TDG values (p = 0.021) (Table 4).
The principal findings of the present study can be summarized as follows: 1) the vessel phantom study found that intraluminal attenuation declined as vessel diameter decreased even though all vessels were filled with the same concentration of contrast media without any active flow; 2) according to the animal study, even in vivo, the peak attenuation of distal small coronary arteries did not reach that of larger arteries; and 3) in the clinical study, TAG showed significant correlation with the diameter gradient, implying that TAG may be a secondary result of decreased intraluminal attenuation because of decreased diameter.
Our study showed that the diameter of the vessel affected intraluminal attenuation. Particularly when the diameter was small, intracoronary luminal attenuation was more greatly affected. For example, in the 2-mm coronary artery model of our phantom study, we observed approximately one-half of the CT number compared with that of the 5-mm coronary artery even though both vessels were filled with the same concentration of contrast media. Although these findings have not previously been reported in the published data, we found 2 examples of supporting evidence from previous studies. First, even in normal coronary arteries, it has been shown that TAG values were different depending on the kind of coronary artery measured. An elaborate, preliminary study by Steigner et al. (7), who analyzed 36 patients with normal coronary arteries using 320-ADCT, explored 3 methods of measuring iodinated contrast opacification gradients (i.e., using the distance from the coronary ostium, lumen cross-sectional area, and lumen short-axis diameter). They found that the contrast gradient over distance was significantly smaller in the right coronary artery compared with the left coronary system despite having normal coronary arteries: –6.5 ± 4.1 for right coronary artery, –13.7 ± 8.0 for LAD, and –12.5 ± 7.8 for LCx. This outcome can be interpreted as that the right coronary artery would be longer and bigger, resulting in a smaller diameter change over distance, compared with the left coronary artery system. Thus, the investigators suggested the superiority of coronary lumen cross-sectional areas and short-axis diameter gradients over distance gradients as these were not significantly influenced by the coronary artery. However, even though the coronary lumen cross-sectional area and short-axis diameter gradients are not influenced by the type of the coronary artery, a smaller diameter or lumen cross-sectional area would still produce a much greater attenuation gradient. Second, similar to our phantom results showing that the slope of the intraluminal attenuation reduction was greater as the diameter became smaller, previous studies (7,8) demonstrated that the diameter range of TAG-positive arteries was much smaller than TAG-negative arteries: from 3 to 1 mm versus from 4.5 to 1.7 mm (7), and from 2.9 to 1.7 mm versus from 4.5 to 2.9 mm (8). Choi et al. (8) also showed that arteries with more severe stenosis had smaller distal caliber. This finding suggests that the downstream coronary arteries with high-grade stenosis or occlusion may become smaller.
The diminution of the attenuation of small objects such as small lung nodules has previously been described as occurring due to the partial volume average effect (16); that is, lung nodules are smaller than the slice thickness. However, the partial volume average effect is not likely to happen in the luminal attenuation of small vessels in our study for 2 reasons. First, in both in vivo and in vitro images, slice thicknesses were smaller than the diameters of vessels, and the size of the ROIs was small enough to avoid containing the vessel wall. Second, in our phantom study, the imaging plane was perpendicular to the vessel orientation, and the vessel was oriented straight; therefore, the partial volume averaging would not have affected the attenuation drop. Rather, we believe that the diminution of intraluminal attenuation in small vessels can be caused by the point-spread function, which is an essential parameter in CT reconstruction algorithms (14). The point-spread function is regarded as a resolution-limiting factor that can lead to images which are not exact replicas of real objects (14,17). Theoretically, CT scanners can produce isotropic image voxels, allowing a uniform sampling of the point-spread function (18), but any deviation from this ideal function causes the point-spread function to widen, resulting in image blurring (14). This effect increases much more in the case of smaller voxels in CT images; thus, in the small vessels of our study, it caused a decrease in CT number.
The animal study revealed that the length of the coronary artery was shown to be an important factor in calculating TAG. For instance, because LAD was shorter than LCx in the dogs studied, the diameter of LAD rapidly decreased from proximal to distal, indicating a higher value of TDG. Thus, a normal LAD exhibited higher TAG than that of a normal LCx even though it did not show a statistical significance. In addition, even normal vessels that are short and show rapid changes in diameter exhibited higher TAG values. The significant correlation between TAG and TDG both in normal vessels and the stenosis models also support our observation.
In the occlusion model, the distal segment to the occlusion exhibited higher attenuation than the proximal segment in the delayed phase 4 heartbeats after the attenuation of the ascending aorta reached its peak. This phenomenon is referred to as the “reversed attenuation sign” representing the retrograde collateral flow distal to the occlusive lesion. In the study by Li et al. (19), they found that the reverse attenuation sign was highly specific for chronic total occlusion and helped to differentiate chronic total occlusion from subtotal occlusion and in diagnosing stent occlusion (20). Choi et al. (21) also used the reverse attenuation sign to identify the presence of well-developed collateral in coronary arteries with chronic total occlusion.
As for our clinical study, we failed to find significant TAG differences between stenotic and nonstenotic coronary arteries and recorded consistent correlations between TAG and TDG in both stenotic and nonstenotic coronary arteries. Debate remains as to the significance of TAG. Several researchers have previously reported a significant difference in TAG between functional significant and nonsignificant stenosis (12), whereas others were not able to find a meaningful difference in TAG in stenoses (22,23). A distinguishing characteristic of our experimental and clinical observations was that it was evident that the attenuation of the coronary artery was highly affected by artery diameter; however, this factor was not considered in the previous research on TAG. Because the diameter of the coronary artery can be one of the biggest confounding factors in evaluating TAG, we emphasize the need to take this diameter into account when assessing TAG as a clinical parameter in CAD. We also want to point out that the TAG of coronary CTA can only represent the rest flow of the coronary artery, as most functional parameters showed positive results only in the stress condition, not in the rest condition. Therefore, it is also hard to expect that the TAG derived from rest phase coronary CTA can be a meaningful functional parameter of coronary artery flow.
First, in the animal study, intracoronary attenuation may have been influenced by motion artifacts, although the canines’ heart rates were controlled with use of beta-blockers. Second, we did not confirm the hemodynamic significance of the coronary stenosis by using hemodynamic data (e.g., flow measurements or FFR in the animal study as well as in the clinical study). Third, in the clinical study, calcium at small coronary artery walls could not have been excluded from the analysis with certainty when the attenuation was below the visual threshold for detection. Lastly, this clinical study is a retrospective, single-center study. Therefore, selection bias may have been present.
We showed that intraluminal attenuation decreased along with diminution of the vessel diameter and that TAG showed a significant correlation with TDG. These findings imply that TAG may be a secondary result because of differences in coronary artery diameters.
COMPETENCY IN MEDICAL KNOWLEDGE: TAG, which is known as a functional CT marker for determining a hemodynamic significance of coronary arterial disease, may be a secondary result because of differences in coronary artery diameters. Our 3-part study (phantom, animal, and clinical) revealed that intraluminal attenuation decreased along with diminution of the vessel diameter, and TAG showed a significant correlation with TDG.
TRANSLATIONAL OUTLOOK: Additional studies are needed to determine a significant relationship between TAG and TDG using hemodynamic data (e.g., flow measurements or FFR). Further research can be conducted to explore if TDG could be used as a potential functional CT marker for determining a significant functional stenosis in the coronary arteries.
For a supplemental table and figures, please see the online version of this article.
This study was supported by grant 04-2014-0540 from the SNUH Research Fund. The authors have reported that they have no relationships relevant to the contents of this paper to disclose.
- Abbreviations and Acronyms
- area detector computed tomography
- coronary artery disease
- computed tomography
- computed tomography angiography
- fractional flow reserve
- left anterior descending coronary artery
- left circumflex coronary artery
- multi-detector row computed tomography
- region of interest
- transluminal attenuation gradient
- transluminal diameter gradient
- Received August 10, 2015.
- Revision received October 8, 2015.
- Accepted October 15, 2015.
- 2016 American College of Cardiology Foundation
- Go A.S.,
- Mozaffarian D.,
- Roger V.L.,
- et al.
- Choi E.K.,
- Choi S.I.,
- Rivera J.J.,
- et al.
- Meijboom W.B.,
- Van Mieghem C.A.,
- van Pelt N.,
- et al.
- Maurovich-Horvat P.,
- Schlett C.L.,
- Alkadhi H.,
- et al.
- Steigner M.L.,
- Mitsouras D.,
- Whitmore A.G.,
- et al.
- Chow B.J.,
- Kass M.,
- Gagne O.,
- et al.
- Koo B.K.,
- Erglis A.,
- Doh J.H.,
- et al.
- Wong D.T.,
- Ko B.S.,
- Cameron J.D.,
- et al.
- Min J.K.,
- Castellanos J.,
- Siegel R.
- Saba O.I.,
- Hoffman E.A.,
- Reinhardt J.M.
- Lim J.,
- Park E.A.,
- Lee W.,
- Shim H.,
- Chung J.W.
- Choi J.H.,
- Kim E.K.,
- Kim S.M.,
- et al.
- Choi J.H.,
- Koo B.K.,
- Yoon Y.E.,
- et al.
- Yoon Y.E.,
- Choi J.H.,
- Kim J.H.,
- et al.