Accuracy of Dual-Source computed tomography in the assessment of plaque morphology  

Study population

From September 2006 to March 2007 we studied 100 patients that were scheduled for conventional coronary angiography (ICA) due to suspected coronary artery disease (CAD) or suspected progression of known CAD. All DSCT studies were performed the day prior to ICA. Exclusion criteria were renal insufficiency (serum creatinine>1.5 mg/dl), hyperthyroidism (basal TSH<0.03μl/l in combination with elevated thyroid hormone levels in the peripheral blood), known allergic reaction against iodinated contrast media or inability to follow breath-hold commands. Five patients were taking a beta-blocker as part of their baseline medication. Additional beta-blocker medication was not administered.

In 13 of these patients (12 men and 1 women; 65±7 years), IVUS was performed in at least one vessels to either evaluate the extent of in–stent restenosis (n=9) or to assess for plaque burden and composition in vessels without angiographic evidence of significant stenosis (n=4).

The study protocol was approved by the local ethics committee, and all patients gave informed consent for participation in the study.

Grey-scale and IVUS-VH imaging protocol and data analysis

For grey-scale and IVUS-VH a phased-array, 20MHz, 3.2Fr IVUS catheter (Eagle Eye, Volcano Corporation, Rancho Cordova, CA) was placed in the distal part of the coronary, distally to the lesion of interest. A motorized pull-back was performed with a pull-back rate of 0.5mm/sec. The pull-back was stopped as soon as the IVUS-catheter reached the guiding catheter. During the pull-back, a grey-scale IVUS was recorded and raw RF data were captured at the top of the R wave. A colour-coded map was reconstructed automatically by a IVUS-VH data recorder (In-Vision Gold, Volcano Corporation, Rancho Cordova, CA).

The volumetric reconstruction of the plaques was performed off-line using a commercial software tool (pcVH 2.1 software, Volcano Corporation, Rancho Cordova, CA). For each segment, both the external elastic membrane and the luminal cross sectional areas were manually defined in accordance to the IVUS interpretation guidelines of the American Heart Association (23). Areas or volumes of the four different IVUS-VH plaque components were automatically calculated for every recorded frame and for the entire imaged segment. Percent vessel stenosis was determined by dividing luminal diameter through vessel diameter for every frame and the maximal percentage was used for further analysis.

CT coronary angiography

All CT scans were performed on a Dual-Source CT scanner (Somatom Definition, Siemens Medical Solutions, Forchheim, Germany). Prior to acquisition of the topogram, patients received a single dose of 0.8 mg glycerol trinitrate.

For contrast enhanced scans, vessel opacification was achieved through automated injection by a power injector (CT2TM, Medtron, Saarbrücken, Germany) of 80ml iomeprol (Imeron® 400, Altana, Konstanz, Germany) at a flow rate of 5ml/s plus a 60ml chaser bolus. Estimation of individual circulation time was based on the test bolus technique, using a 20ml bolus and dynamic evaluation software (Dyn EvaTM, Syngo®, Siemens, Forchheim, Germany).

Collimation was 32 x 0.6mm, slice acquisition 64 x 0.6mm using the z-flying focal spot technique, gantry rotation time 330ms, pitch 0.20-0.43 adapted to heart rate, tube voltage 120 kV and maximum tube current 400 mAs per rotation. For dose reduction, prospective tube current modulation was applied. Thereby, at heart rates below 60 bpm, full tube current was applied from 60-70%, at 60-70 bpm from 50-80%, and at heart rates above 70 bpm from 30 to 80% of the cardiac cycle.

For data reconstruction, a single-segment reconstruction algorithm was applied which uses the data of a quarter rotation of both detectors for image reconstruction.
An initial reconstruction window was based on the results of a test series which was obtained in a transverse plane at the level of segment 2 and which displayed reconstruction window offsets by 5% of the entire cardiac cycle. In case of motion artefacts in the initial reconstruction, further reconstructions were obtained in 5% increments of the cardiac cycle until all individual arteries could be visualized at optimal image quality.

Effective slice thickness was 0.75mm with a reconstruction increment of 0.4mm. Data sets were filtered with a medium-soft convolution kernel (B26f).

DSCT image analysis

CT data were referred to an offline workstation (Leonardo, Siemens, Germany) and assessed using the Circulation III package (Siemens, Germany) and the Plaque Lens (PL) software (Siemens, Germany) for vessel or plaque analysis. The software is based on curved multiplanar reconstructions (CPRs) and images are displayed along and orthogonally to the centre-line of coronary arteries. For additional orientation, thin-slab maximum-intensity-projections and 3-D volume renderings were used.

Evaluation of plaque was performed according to the following multi-step protocol, illustrated in Figure 1. First, reader 1 empirically compared IVUS-VH and CT-PL images to assess the optimal image display settings for identical edge definition of the outer elastic membrane region in both modalities. Our approach modelled the technique which has previously been suggested by Leber et al. (18). Accordingly, 5 patients were randomly selected and in each patient 2 segments defined in CT and IVUS by both fiduciary markers (side branches or stents) and segment length. Thereby, proximal and distal segment boundaries were manually defined on CPRs using a lineal measurement function. A discrepancy of less than 1 mm between IVUS and CT segment length was tolerated. The automatic vessel edge definition of the PL-software was employed and, whenever necessary, the vessel boundaries subsequently modified to achieve a close match (± 10%) to the vessel volumes in IVUS. Window width and level were then manipulated until the vessel edge defining line in the PL-software most closely equalled the interface of enhancing vessel and non-enhancing wall or peri-vascular structures. Widths and levels were recorded and set in relation to the mean HU density within the lumen of the distinct segment (mean of three ROI (region of interest) measurements). In doing so, we determined 212% or 66% of luminal CT density as the optimal average window widths and level settings for further analyses.

In a second step, reader 2 identified the 20 plaque loaded vessel segments that had previously been defined in IVUS according to segment length and fiduciary markers. Again, a discrepancy of 1 mm was accepted with regard to segment length. Whenever necessary, automatic segmentation of vessel edge was manually modified according to the above criteria.

For comparison with IVUS, maximal percent vessel stenosis was determined by manually tracing the luminal cross sectional area on the axial slice visually determined as showing the narrowest luminal diameter of the lesion.

For all segments, volumes of vessel, lumen as well as low, medium and high density plaque were automatically displayed by the PL-software. Thereby, default HU-ranges are 0-70 for low density (dark green), 71-130 for medium density (light green) and 401- for high density plaque (purple). The vessel lumen is defined by a HU range of 131-400 (yellow).

In a third step, reader 2 and 3 interactively manipulated the default HU-ranges of the PL-software until HU-based colours maps most closely matched the visual aspect of plaque morphology. In doing so, edge definition of the vessel segment remained unchanged.
Subsequently, reader 2 was de-blinded of the IVUS-VH results and adjusted the HU-ranges of the analysis software so that overall plaque volume and volume of the individual plaque components would most closely match the IVUS results (± 5%). Thereby, the volumes of IVUS-VH fatty-fibrous plaque and necrotic core were summed up to equal the low density plaque component in the PL-software. Resulting HU cut-off values were recorded.

In a next step, the mean of these cut-off values was applied to all plaques, again keeping the vessel edge definition unchanged (reader 2).

Finally, optimal, individual HU-ranges as obtained in direct comparison to IVUS-VH were re-applied to each individual plaque and vessel edges now circumferentially expanded into the epi-vascular fat. On axial MPRs, the edge defining line was then re-adjusted so as to precisely match the outside edge of the colour maps (reader 2).

For assessment of inter-observer variability, the definition of the outer vessel boundaries was eventually repeated by reader 3 using identical proximal and distal segment markers. For the inter-observer comparison of plaque and vessel volumes HU-ranges were set to the mean of IVUS correlated thresholds.

Statistical Analyis

Statistical analysis was performed with software (JMP version 6, SAS Institute, Cary, NC; GraphPad Prism version 4.00, San Diego; SPSS version 15, Chicago). A P value of less than 0.05 indicated statistical significance.

Continuous variables are expressed as mean±standard deviation and comparisons between volumes were performed using either t-test or ANOVA plus Tukey`s post test for paired observations. Bland-Altman analysis is used to display the systematic error and confidence interval (CI) between the measurements whilst the correlation between CT and IVUS variables is assessed by calculation of the Pearson`s correlation coefficient. The impact of vessel attenuation on HU cut-off values for lumen or calcified plaque is evaluated by linear regression. Interobserver reliability was assessed by calculating the 2-way random sinlge-measure intraclass correlation coefficient (ICC).

Results

DSCT and IVUS were successfully performed in all patients without complications. Mean heart rate was 64±8 bpm (range 50 to 78 bpm) and good image quality was obtained in all CT studies. Plaque loaded segments as defined by IVUS were unequivocally identified on CT CPRs through fiduciary markers and segment length. The distribution of lesions was the following: 3 lesion in the left main artery; 8 lesions in the left anterior descending (n=4 segment 6; n=3 in segment 7; n=1 in segment 8), 6 lesions in the circumflex (n=5 segment 11; n=1 segment 12) and 3 in the right coronary artery (n=2 segment 1; n=1 segment 1).

Default HU-ranges for low density, medium density, lumen and high density plaque were 0-70, 71-130, 131-400 and 401+. (Wiederholung, oder?)The corresponding mean HU-ranges resulting from visual adjustment or IVUS correlation of volumes were -15-90, 91-174, 175-471, and 472+ or -6-66, 67-153, 154-446, and 447+.

With both visually adjusted and IVUS correlated HU settings, cut-off values for calcified plaques were significantly correlated with luminal attenuation. Likewise, linear regression showed significant correlation between vessel enhancement and the IVUS correlated cut-offs for lumen (Figure 2).

Mean percent vessel stenosis as determined by CT (49±12%) was not significantly different from IVUS (51±14; P=0.22) and showed a close correlation (r2=0.71).

The volumes of vessel, lumen, plaque and plaque components in IVUS-VH versus DSCT with variable HU-ranges or vessel edge definitions are provided in Table 1. Bland-Altman analysis revealed considerable biases with all CT settings except the use of mean IVUS adjusted HU-thresholds (Table 1; Figure 3). However, significant overestimation was only observed with use of default settings for calcified plaque, use of visual adjustment for fatty plaque or modified vessel edge definition for vessel, plaque and non-calcified components. Significant underestimation was present with use of default settings for fibrous plaque and use of visual adjustment for lumen.

Spearman`s correlation coefficients for vessel, lumen and plaque volumes are provided in Table 1. Strongest association for total plaque volume was obtained using the mean of IVUS adjusted HU cut-offs (Figure 4). The same was true regarding best correlation for fatty-fibrous or fibrous plaque components. A visual HU-range definition proved superior in quantifying the high density plaque (Table 1).

Intraclass correlation coefficients for volumes of vessel, lumen, plaque and low, medium or high attenuation components were 0.96, 0.99, 0.94, 0.85, 0.97 and 1.00 suggesting a small inter-observer variability.

Discussion

Only recently, spectrum analysis of IVUS-derived radiofrequency has emerged as an attractive method to not only assess total plaque burden but also accurately differentiate the various components and, thereby, produce a virtual histology of plaque morphology. However, due to numerous restrictions IVUS-VH is rarely used in clinical practice. On the other hand, MSCT has repeatedly demonstrated its potential to identify and classify coronary plaque. Multiple studies have proved significant differences between CT density units for hypoechoic, hyperechoic and calcified lesions, suggesting that HU is reflecting plaque composition (17,21,24-27). However, the accuracy of MDCT in unveiling plaque morphology has been limited by partial volume and motion artefacts. Also, in most studies, CT density values have not primarily een employed in the assessment of plaque morphology but have only retrospectively been obtained from visually detected plaque components. Moreover, plaque burden has usually been addressed in terms of cross-sectional areas only; the volume of plaque components has rarely been subject of systematic evaluation in cardiac CT. Lately, new software has been designed which segments and quantifies different plaque components by HU-tresholds and, thereby, provides colour maps of volumes just like virtual histology in IVUS-derived radiofrequency.

This is the first study to assess DSCT and advanced analysis software regarding their potential to provide a CT virtual histology, using IVUS-derived radiofrequency as a standard of reference.

Primary Findings

Our primary findings are threefold: First, prospective use of a HU range based analysis provides moderate or good ?correlation of volumes for total plaque burden as well as its components with IVUS-VH. Secondly, whilst fixed HU cut-offs prove superior accuracy in determining the volumes of non-calcified plaque, visual modification of HU-ranges is advantageous for hard plaque quantification. Thirdly, optimal HU-ranges as observed in our study partly differ from cut-off values reported in many previous investigations. Best HU thresholds for coronary lumen, thereby, significantly correlate with vessel attenuation.

In comparison to prior studies by Moselewski or Achenbach et al., we found a considerably closer correlation between CT and IVUS for percent vessel stenosis (15,19). A plausible explanation for is that our analysis was limited to clearly defined and visually detectable plaques and did not include false positive or negative findings. On the other hand, spatial as well as temporal resolution of DSCT is much higher than that of MSCT used in the above mentioned studies.

There is a considerable discrepancy in literature regarding the concordance of plaque burden in MSCT and IVUS. In fact, correlation coefficients for plaque burden as assessed in terms of cross sectional area show a wide variation from r=0.55 to r2=0.82 (18,19,28). Interestingly, the degree of correlation as determined in our study is in good agreement with results from Achenbach or Leber et al. who, likewise, quantify plaque in terms of volume (15,16). However, in contrast to their findings we did not find a significant bias for total plaque or softplaque (besser non-calcified) components.

Likewise, use of a HU-based analysis resulted in a considerably improved inter-observer reproducibility. Reasons for a reduction of inter-observer variability with the HU-based approach include elimination of manual plaque segmentation. Moreover, results of HU-based plaque analysis are largely independent of display settings and merely correlate with modifications in the semiautomatic definition of vessel boundaries. Notably, in a head-to-head comparison within our investigation, visual adjustment of colour maps to non-calcified plaque was associated with inferior correlation to IVUS.

In our feasibility study, we employed the improved temporal resolution of DSCT to achieve a reduction of motion and secondary partial volume artefacts (29). In combination with our analysis method, such technical advance likely promoted the feasibility to accurately quantify not only the total plaque load but also its major components.

However, despite promising results in our pilot study, a number of major limitations have been observed regarding prospective HU-based plaque characterization. First, the method is highly sensitive to modification of thresholds; even limited changes, such as between the default settings and the mean of IVUS correlated HU cut-offs, lead to significant differences in volumes of plaque components.

Secondly, even with optimized HU-ranges, correlation of non-calcified plaque to IVUS was only moderate. The primary reason is the limited spatial resolution of CT which is certainly still insufficient to break down the complex plaque morphology. In fact, the latter is featuring lipid, fibrous and calcified components not only side by side but more often interlaced into each other. Hence, partial volume effects most likely account for the overlap of density values that have repeatedly been demonstrated for fatty or fibrous plaque components (17,21,30). In addition to the issue of partial volume averaging, CT density of fatty-fibrous plaque and the necrotic core are most probably not significantly different and, thus, the two tissue types turn out to be inseparable in CT-VH.

In our study, the sum of fatty-fibrous plaque and necrotic core was, therefore, used for comparison with low density plaque in CT. While this inability to separate the necrotic core (at least in case of a small dimension of the necrotic core) represents a considerable drawback, the potential of MSCT to discriminate fibrous from fatty plaque components remains crucial to a non-invasive definition of the vulnerable plaque. On the other hand, it should be noted, that IVUS-VH is, equally, subject to methodical limitations. In fact, in multiple subjects, we found IVUS to project necrotic core areas into regions of hard plaque induced acoustic shadow. Insofar, misinterpretation in the reference method may have had a small role in the moderate degree of correlation.
In previous studies, CT density values of non-calcified plaque have frequently been obtained from ex vivo autopsy material or from ROI measurements in tissue that was visually correlated to hypoechoic, hyperechoic areas in IVUS. We assumed that assessment of average attenuation would be more accurate in vivo and when obtained from the entire plaque components the volumes of which were previously matched to IVUS-VH. In doing so, we found markedly higher HU cut-offs for fibrous plaque than earlier reported in literature (17,21,30). Interestingly, our results are in good agreement with more recent comparisons of MSCT to IVUS (24-26).

Arguably, differences in radiation dose application, detector configuration and contrast protocols do have an effect on optimal thresholds for tissue with CT density values close to that of the vessel lumen. This notion is in accordance with the moderate, yet significant correlation we found between vessel attenuation and optimal HU cut-offs for fibrous plaque. An even stronger association was evident between CT density values in the vessel and best HU threshold for calcification. This very association is causative to the poor correlation between volumes of calcified plaque and IVUS whenever fixed HU-ranges were applied. Indeed, visual adjustment of HU cut-offs ensured best match for calcium.

In summary, our findings are, thus, suggesting a flexible approach to prospective HU-based plaque analysis: whilst fixed thresholds are advantageous for soft plaque (non calcified) quantification, visual adjustment is advisable to optimize outcome for calcified lesions. Unquestionably, the latter will, however, best be quantified by obtaining the calcium volume score from native CT scans. In the long run, software might be improved to automatically adjust thresholds to variable vessel attenuation.

It is important to note that the vessel edge definition was only semi-automatic in our study and that manual corrections were performed to ensure a perfect match to the outer elastic membrane interface. In a final step of our experiment, vessel edge definition was achieved by using the HU thresholds for fatty plaque as obtained in our head to head calibration with IVUS. Not surprisingly, such adjustment of vessel edge led to significant increase of vessel volume and systematic overestimation of non-calcified plaque. This is highlighting the fundamental difference in what IVUS and CT are ultimately measuring: whilst IVUS merely accesses the intraluminal plaque and media, CT visualizes the entire atheroma, including its adventitial component. The latter may be substantial which is especially true in advanced disease (31,32). Nevertheless, even with this potential to measure the entire atherosclerotic process, CT cannot separate all its components and will above all not discriminate non-calcified plaque and vessel wall. This results in a systematic overestimation of plaque volume whenever the vessel edge boundaries are moved beyond the outer elastic membrane.

The small number of observations is an obvious limitation to our pilot study. Since IVUS was performed in proximal lesion only, vessel diameters were relatively large. The same was true for average plaque volumes. Reproducibility of our results in more distal segments and with smaller plaques needs confirmation in further studies (ist aber schwierig, weil man den IVUS-Katheter gar nicht bis in die Peripherie kleiner Gefäß bekommt). The potential to use two X-ray tubes with different voltages to thus further characterize plaque composition has not been exploited.

1. Falk E, Shah PK, Fuster V. Coronary plaque disruption. Circulation 1995;92:657-71.
2. Kragel AH, Reddy SG, Wittes JT, Roberts WC. Morphometric analysis of the composition of atherosclerotic plaques in the four major epicardial coronary arteries in acute myocardial infarction and in sudden coronary death. Circulation 1989;80:1747-56.
3. Libby P. Molecular bases of the acute coronary syndromes. Circulation 1995;91:2844-50.
4. Mintz GS, Painter JA, Pichard AD, et al. Atherosclerosis in angiographically "normal" coronary artery reference segments: an intravascular ultrasound study with clinical correlations. J Am Coll Cardiol 1995;25:1479-85.
5. Nissen SE, Gurley JC, Grines CL, et al. Intravascular ultrasound assessment of lumen size and wall morphology in normal subjects and patients with coronary artery disease. Circulation 1991;84:1087-99.
6. Topol EJ, Nissen SE. Our preoccupation with coronary luminology. The dissociation between clinical and angiographic findings in ischemic heart disease. Circulation 1995;92:2333-42.
7. Nishimura RA, Edwards WD, Warnes CA, et al. Intravascular ultrasound imaging: in vitro validation and pathologic correlation. J Am Coll Cardiol 1990;16:145-54.
8. Potkin BN, Bartorelli AL, Gessert JM, et al. Coronary artery imaging with intravascular high-frequency ultrasound. Circulation 1990;81:1575-85.
9. Yock PG, Linker DT. Intravascular ultrasound. Looking below the surface of vascular disease. Circulation 1990;81:1715-8.
10. Moore MP, Spencer T, Salter DM, et al. Characterisation of coronary atherosclerotic morphology by spectral analysis of radiofrequency signal: in vitro intravascular ultrasound study with histological and radiological validation. Heart 1998;79:459-67.
11. Nair A, Kuban BD, Tuzcu EM, Schoenhagen P, Nissen SE, Vince DG. Coronary plaque classification with intravascular ultrasound radiofrequency data analysis. Circulation 2002;106:2200-6.
12. Nasu K, Tsuchikane E, Katoh O, et al. Accuracy of in vivo coronary plaque morphology assessment: a validation study of in vivo virtual histology compared with in vitro histopathology. J Am Coll Cardiol 2006;47:2405-12.
13. Rumberger JA. Noninvasive coronary angiography using computed tomography: ready to kick it up another notch? Circulation 2002;106:2036-8.
14. Schoepf UJ, Becker CR, Ohnesorge BM, Yucel EK. CT of coronary artery disease. Radiology 2004;232:18-37.
15. Achenbach S, Moselewski F, Ropers D, et al. Detection of calcified and noncalcified coronary atherosclerotic plaque by contrast-enhanced, submillimeter multidetector spiral computed tomography: a segment-based comparison with intravascular ultrasound. Circulation 2004;109:14-7.
16. Leber AW, Becker A, Knez A, et al. Accuracy of 64-slice computed tomography to classify and quantify plaque volumes in the proximal coronary system: a comparative study using intravascular ultrasound. J Am Coll Cardiol 2006;47:672-7.
17. Leber AW, Knez A, Becker A, et al. Accuracy of multidetector spiral computed tomography in identifying and differentiating the composition of coronary atherosclerotic plaques: a comparative study with intracoronary ultrasound. J Am Coll Cardiol 2004;43:1241-7.
18. Leber AW, Knez A, von Ziegler F, et al. Quantification of obstructive and nonobstructive coronary lesions by 64-slice computed tomography: a comparative study with quantitative coronary angiography and intravascular ultrasound. J Am Coll Cardiol 2005;46:147-54.
19. Moselewski F, Ropers D, Pohle K, et al. Comparison of measurement of cross-sectional coronary atherosclerotic plaque and vessel areas by 16-slice multidetector computed tomography versus intravascular ultrasound. Am J Cardiol 2004;94:1294-7.
20. Schoenhagen P, Tuzcu EM, Stillman AE, et al. Non-invasive assessment of plaque morphology and remodeling in mildly stenotic coronary segments: comparison of 16-slice computed tomography and intravascular ultrasound. Coron Artery Dis 2003;14:459-62.
21. Schroeder S, Kopp AF, Baumbach A, et al. Noninvasive detection and evaluation of atherosclerotic coronary plaques with multislice computed tomography. J Am Coll Cardiol 2001;37:1430-5.
22. Flohr TG, McCollough CH, Bruder H, et al. First performance evaluation of a dual-source CT (DSCT) system. Eur Radiol 2006;16:256-68.
23. Mintz GS, Nissen SE, Anderson WD, et al. American College of Cardiology Clinical Expert Consensus Document on Standards for Acquisition, Measurement and Reporting of Intravascular Ultrasound Studies (IVUS). A report of the American College of Cardiology Task Force on Clinical Expert Consensus Documents. J Am Coll Cardiol 2001;37:1478-92.
24. Carrascosa PM, Capunay CM, Garcia-Merletti P, Carrascosa J, Garcia MF. Characterization of coronary atherosclerotic plaques by multidetector computed tomography. Am J Cardiol 2006;97:598-602.
25. Iriart X, Brunot S, Coste P, et al. Early characterization of atherosclerotic coronary plaques with multidetector computed tomography in patients with acute coronary syndrome : A comparative study with intravascular ultrasound. Eur Radiol 2007.
26. Pohle K, Achenbach S, Macneill B, et al. Characterization of non-calcified coronary atherosclerotic plaque by multi-detector row CT: comparison to IVUS. Atherosclerosis 2007;190:174-80.
27. Schmid M, Pflederer T, Jang IK, et al. Relationship between degree of remodeling and CT attenuation of plaque in coronary atherosclerotic lesions: An in-vivo analysis by multi-detector computed tomography. Atherosclerosis 2007.
28. Hara T, Yamada S, Hayashi T, et al. Accuracy of nonstenotic coronary atherosclerosis assessment by multi-detector computed tomography. Circ J 2007;71:911-4.
29. Reimann AJ, Rinck D, Birinci-Aydogan A, et al. Dual-source computed tomography: advances of improved temporal resolution in coronary plaque imaging. Invest Radiol 2007;42:196-203.
30. Becker CR, Nikolaou K, Muders M, et al. Ex vivo coronary atherosclerotic plaque characterization with multi-detector-row CT. Eur Radiol 2003;13:2094-8.
31. Fayad ZA, Fuster V, Fallon JT, et al. Noninvasive in vivo human coronary artery lumen and wall imaging using black-blood magnetic resonance imaging. Circulation 2000;102:506-10.
32. Gradus-Pizlo I, Bigelow B, Mahomed Y, Sawada SG, Rieger K, Feigenbaum H. Left anterior descending coronary artery wall thickness measured by high-frequency transthoracic and epicardial echocardiography includes adventitia. Am J Cardiol 2003;91:27-32.