- Email: [email protected]
Impact of vessel curvature on the accuracy of three-dimensional intravascular ultrasound: Validation by phantoms and coronary segments
Impact of Vessel Curvature on the Accuracy of Three-Dimensional Intravascular Ultrasound: Validation by Phantoms and Coronary Segments Peter Sta¨hr, MD, Thomas Voigtla¨nder, MD, Hans-Ju¨rgen Rupprecht, MD, Patrick Aschenbru¨cker, BS, Hayraet Mamtimin, MD, Ru¨diger Brennecke, PhD, Mike Otto, MD, Peter J. Fitzgerald, MD, PhD, and Ju¨rgen Meyer, MD, Mainz, Germany, and Stanford, California
Background: Three-dimensional intravascular ultrasound (IVUS) is used for volumetric assessment of arteriosclerotic plaque burden and restenotic tissue at follow-up after coronary interventions. However, the accuracy of these measurements, especially in tortuous vessels, is unclear. Methods: A commercially available electrocardiogram (ECG)-gated 3-dimensional-IVUS system was tested in volume-validated straight and curved hydrocolloid phantoms and in volume-validated coronary specimens. Catheter withdrawal (30 MHz, 3.2F) was triggered using standardized ECG source with 0.2-mm step intervals per cardiac cycle simulation.
A
t present, 3-dimensional (3D)-intravascular ultrasound (IVUS) is used as an essential research tool in the evaluation of specific antirestenosis therapies.1-3 It may also serve as a valuable tool in the understanding and assessment of the regression/progression of arteriosclerosis over time in a given patient population. For example, a 3D-IVUS investigation has been initiated to evaluate the response of arteriosclerosis to treatment with cerivastatin sodium, nifedipine, or both.4 In contrast, quantitative coronary angiography has known limitations in demonstrating small differences in vessel architecture, as seen in secondary prevention trials.5,6 In addition, semiautomated 3D-IVUS systems with electrocardiogram (ECG) triggering for the detection of the intimal leading edge and the external From the 2nd Medical Clinic (Cardiology) and Department of Pathology (M.O.), Johannes Gutenberg-University, Mainz, Germany; and Stanford University Medical School (P.S., P.J.F.), California. Supported by a Feodor-Lynen grant from the Alexander von Humboldt-Stiftung (P.S.). Reprint requests: Peter J. Fitzgerald, MD, PhD, Stanford University School of Medicine, Division of Cardiovascular Medicine, 300 Pasteur Dr, Room H3554, Stanford, CA 94305 (E-mail: [email protected]). Copyright 2002 by the American Society of Echocardiography. 0894-7317/2002/$35.00 ⫹ 0 27/1/120700 doi:10.1067/mje.2002.120700
Results: On the basis of automated phantom volume measurements, IVUS overestimated true phantom volume (relative error ⴝ [measured V ⴚ true V]/true V ⴛ 100) by a median of 0.9%, 0.25%, and 1.96% for straight, mildly curved, and severely curved segments, respectively. The true volume of the coronary specimens was overestimated by a median of 5.79%. Conclusion: A median percentage deviation of 3-dimensional-IVUS-measured volumes from the true volumes of less than 10% in phantoms and coronary artery segments can be achieved. (J Am Soc Echocardiogr 2002;15:823-30.)
vessel contour (media-adventitia boundary) have been developed for volume quantitation.7-9 However, with all these volume measurement techniques, curvatures may influence the IVUS-quantitation of vessel lumen and plaque volume.10 The aim of this study was to identify volume variations in ECG-gated 3D-IVUS-measurements in both straight and curved hydrocolloid phantoms and in irregularly shaped coronary artery specimens. These validations may contribute to the accuracy of future research studies investigating volumetric changes in arteriosclerotic plaque over time and the extent of restenotic tissue proliferation.
METHODS IVUS Imaging and ECG-gated 3D-IVUS Image Acquisition and Analysis Imaging of the phantoms and coronary artery segments was performed in vitro using a routinely available mechanical IVUS system (ClearView cardiovascular imaging system, Boston Scientific Corp, Maple Grove, Minn) with a 30 MHz catheter (3.2F Ultracross and Scimed, Boston Scientific Corp, Maple Grove, Minn). The catheter used was sheath based, meaning that the transducer element could be moved in a transparent trajectory.
823
Journal of the American Society of Echocardiography August 2002
824 Sta¨hr et al
Figure 1 Longitudinally reconstructed intravascular ultrasound images of 3 different phantom types (A, straight; B, mildly curved; C, severely curved) and D, coronary specimen. Note almost straight shape of curved phantom images (B, C) and catheter touching wall. Table 1 Absolute values of phantom dimensions (4 phantoms of each type with 10 IVUS measurements in each phantom)
Phantom type (n ⴝ 4 each)
Straight P Mildly CP Severely CP
True lumen diameter (mm)
True central lumen length (mm)
True volume (mean ⴞ SD) (mm3)
Measured volume (mean ⴞ SD) (mm3)
Absolute difference (mean ⴞ SD) (mm3)
2.7 2.7 2.7
20 20 20
279.02 ⫾ 2.01 257.20 ⫾ 2.17 264.67 ⫾ 3.52
280.52 ⫾ 8.85 257.55 ⫾ 11.1 270.64 ⫾ 1.44
1.5 ⫾ 8.24 0.36 ⫾ 10.99 5.97 ⫾ 9.64
IVUS, Intravascular ultrasound; C, curved; P, phantom.
The ECG-gated 3D-IVUS image acquisition and digitization (TomTec Imaging Systems Inc, Unterschleissheim, Germany) has been described in detail elsewhere.8 The transducer, when positioned in the phantom or coronary artery segment, was withdrawn by a pullback device in step intervals of 0.2 mm per cardiac cycle, triggered by a standardized ECG source. IVUS images were acquired and digitized 40 ms after the peak of the R-wave. Consecutively, the transducer was withdrawn to acquire the next image. Each set of digitized IVUS images was analyzed off-line with an automated, computerized contour detection algorithm.7,8 Two longitudinal sections were constructed and contours corresponding to the inner lumen border and outer phantom/segment border were automatically identified by a minimal cost algorithm7,8 (Figure 1). If necessary, these longitudinal contours were edited with computer assistance. The longitudinal contours were transformed to individual edge points on the cross-sectional images. On the basis of these edge points, the system automatically suggested the border contour in each cross-sectional image (Figure 1). In cases with incomplete border evaluation, the
contour presented could be manually edited by the operator. The wall cross-sectional area (A ⫽ outer cross-sectional area ⫺ lumen cross-sectional area) was calculated for each planar image. Volume of the phantom/segment was calculated by: V⫽n⌺
i⫽1
(A1 ⫻ H)
where H is the thickness of the slice (represented by a single tomographic IVUS image) and n is the number of IVUS images in the 3D data set.8 Phantom Volume Measurements by IVUS Phantoms used for volume measurements were made from a hydrocolloid (Optiloid, Riss-Dental, Hanau, Germany) with tissue-like ultrasound properties.11 All phantoms had a central lumen length of 20 mm, an inner lumen diameter of 2.7 mm, and an outer diameter of 5 mm (Table 1). Three different types of phantoms were produced: straight, mildly curved (radius 30 mm), and severely curved (radius 10 mm) (Figure 2).
Journal of the American Society of Echocardiography Volume 15 Number 8
Sta¨hr et al
825
Figure 2 Sketch of true dimensions of different phantom types: A, straight; B, mildly curved; C, severely curved.
Within the 3 types, 4 phantoms each were used with 10 pullbacks being performed for each phantom. Volumes were measured off-line using the automated, computerized system described previously. Coronary Artery Segment Volume Measurements by IVUS Ten straight segments of coronary arteries (5 left anterior descending coronary artery, 2 left circumflex coronary artery, 3 right coronary artery) of 5 patients obtained by autopsy were investigated by IVUS after they were removed from the atrioventricular groove or left anterior descending coronary artery groove. Ten pullbacks in each segment were performed. Five of the 10 segment contours derived automatically were optimized manually in the planar images by one of the authors (P.S.). Thus, volume measurements from automated and manually corrected border detection could be compared. Quantitative volume measurements with manual correction were repeated at least 6 weeks apart and used to assess their reliability. The correlation coefficient obtained from linear regression analysis and the percent error obtained by taking the absolute difference divided by the initial measurements were used to express the intraobserver variability. The intraobserver correlation coefficient and percent error were r2 ⫽ 0.99 and 1.2% ⫾ 0.8% (mean ⫾ SD). The sheath-based catheter was centered, without the use of a guide wire, at the inlet and outlet of the phantoms or artery segments. The pullbacks were performed in 20°C water. After each pullback, it was mandatory to supply a reference distance for the 3D software. To provide reliable referencing, the diameter of a hydrocolloid phantom (5 mm) was used for distance calibration. This reference was included at the end of each pullback in both phantoms and artery segments. As a result of this procedure, differences in sound velocities of blood and water, leading to errors in terms of overestimation or underestimation, could be minimized.12
Validation of Phantom and Coronary Segment Volumes The true volume of the phantoms and artery segments was determined by pycnometer measurements. The method is on the basis of weighing (Microgram weighing machine, Sartorius, Goettingen, Germany) the water volume displaced by the body after immersion into the definite volume of the water-filled pycnometer (20°C). The water weight divided by the water density yields the body volume. The accuracy of measurements is within a range of ⫾0.07%. Statistical Analysis Results are given as median with the 25th, 75th, 10th, and 90th percentiles or as mean ⫾ SD. Systematic differences of the relative volume errors between the different kinds of phantoms or arteries were tested using the t test for unpaired observations. When normality or equal variance testing failed, the Mann-Whitney rank sum test was applied. A P value ⬍ .05 was considered to be statistically significant.
RESULTS Phantom Volume Measurements by IVUS Sample images from each phantom type are shown in Figure 2. The true phantom curves are not visible in the longitudinal IVUS images; the curved phantoms almost appear straight. However, because of the curvatures, the catheter is forced toward the wall (Figure 1, B and C). The automatic contour detection matched the phantom borders very well, so that manual contour optimization was not necessary. The absolute true volumes (evaluated by pycnometer measurements) and volumes (measured by IVUS) are listed in Table
Journal of the American Society of Echocardiography August 2002
826 Sta¨hr et al
Figure 3 Three-dimensional–intravascular ultrasound volume measurements: Deviation of measured volume from true phantom or coronary artery segment volume as expressed by relative error (%) ⫽ (measured V ⫺ true V)/true V ⫻ 100. V, Volume. Table 2 Absolute values (after manual contour optimization) of coronary artery specimens (with 5-10 IVUS pullbacks in each segment) Artery no.
Mean measured LD (mm)
Mean measured central LL (mm)
True V (mm3)
Measured V (mean ⴞ SD) (mm3)
Absolute difference (mean ⴞ SD) (mm3)
1 2 3 4 5 6 7 8 9 10
2 2.6 1.5 2.9 2.3 2.5 2.2 1.8 2.3 2.5
19.8 20.8 17.6 22.2 20.2 18 16.5 17.8 19 23
289.35 301.29 137.76 337.61 190.34 210.38 86.16 146.00 258.47 282.83
284.84 ⫾ 8.87 260.69 ⫾ 14.10 146.56 ⫾ 12.43 331.75 ⫾ 8.84 204.05 ⫾ 7.62 221.91 ⫾ 3.98 101.64 ⫾ 3.67 160.03 ⫾ 4.39 270.61 ⫾ 1.58 304.30 ⫾ 2.41
⫺4.51 ⫾ 8.87 ⫺40.60 ⫾ 14.10 8.80 ⫾ 12.43 ⫺5.86 ⫾ 8.84 13.71 ⫾ 7.62 11.53 ⫾ 3.98 15.48 ⫾ 3.67 14.03 ⫾ 4.39 12.14 ⫾ 1.58 21.47 ⫾ 2.41
IVUS, Intravascular ultrasound; LD, lumen diameter; LL, lumen length; V, volume.
1. On the basis of the automated volume measurements, 3D-IVUS overestimated the true volume (relative error ⫽ [measured ⫺ true]/true volume ⫻ 100) by 0.9% (median, 25th to 75th percentiles: ⫺1.15% to 2.45%) in the straight phantoms and by 0.25% (25th to 75th percentiles: ⫺2.21% to 2.6%) in the mildly curved phantoms (P ⫽ .65, Figure 3). In the severely curved segments, the true volume was overestimated by 1.96% (25th to 75th percentiles: ⫺1.15% to 5.27%), which is significantly different from the relative error in both the straight and mildly curved phantoms (P ⫽ .035 and P ⫽ .045, respectively, Figure 3).
Coronary Artery Segment Volume Measurements by IVUS Table 2 lists dimensions and volumes of the artery segments measured by IVUS (after manual optimizing of contour delineation). IVUS measurements overestimated true segment volumes, evaluated by pycnometer measurements, by a median relative error of 5.79% (25th to 75th percentiles: ⫺0.01% to 9.23%) (Figure 3). Compared with the error observed in the phantom measurements, the relative error in the coronary artery segments demonstrated a wider variability (wider 10th, 25th, 75th, and 90th percentiles, Figure 3).
Journal of the American Society of Echocardiography Volume 15 Number 8
Sta¨hr et al
827
IVUS measurements on the basis of automatic contour delineation alone, without manual optimization, showed a higher relative error (median) of 8.1% (25th to 75th percentiles: 3.37% to 14.63%), P ⫽ .018.
sectional area in curved phantoms, reported an overestimation of the true cross-sectional area by up to 18% in higher curvatures. The differences in errors may be partly because of different distances between the transducer and the vessel wall or plaque.14
DISCUSSION
Volume Measurements in Coronary Artery Segments
Phantom Volume Measurements 3D-IVUS overestimated the true volume of the straight and mildly curved phantoms (curve radius 30 mm) with a median of up to 0.9% (Figure 3) and not higher than 1.96% in the severely curved segments (curve radius 10 mm). The higher overestimation in severely curved segments was because of geometric factors: compression/expansion of the vessel and oblique cross-sectional areas within the curvature.10,13 Similar to our results, luminal volume measured in a straight paraffin tubular phantom by von Birgelen et al,9 using the same 3D-system without ECG gating, was overestimated by 0.25% to 1.72%. It should be stressed that von Birgelen’s investigation was aimed at the lumen volume in contrast to the total phantom volume in our investigation. For measuring the phantom volume, a contour detection of the inner and outer border is necessary, with the potential of increased systematic error. In this study, we used a sheath-based catheter with the catheter moved inside a transparent trajectory. The transducer traverses along a path that is relatively straight compared with the local vessel curvature.14 According to Schuurbiers et al,14 the volume error with this system was primarily dependent on the curvature of the catheter pullback trajectory; vessel tortuosity itself played a role only in so far as it affected the local curvature of the pullback trajectory. Furthermore, with increasing curvature of the pullback trajectory, the volume error was increasing, which is in line with our study. Hence, the error in the same curve may be even higher when using a nonsheath-based catheter system in which the catheter is moved as a whole in the vessel. Other groups have described an overestimation of the lumen volume or cross-sectional area of curved segments: Wiet et al15 found an overestimation of the true lumen volume in curved tubes (length 25 mm) of 1% (for a curve radius of 100 mm) and 35% (for a curve radius of 20 mm) measured with another 3D-IVUS system. The group recommended only lower curvatures (curve radius ⬎ 60 mm) for 3D-IVUS-measurements. In our study, even in segments with severe curvature (radius of 10 mm), we found an overestimation of less than 10%. In addition, Schwarzacher et al,13 measuring the cross-
In addition to phantoms, we examined the variations of 3D-IVUS measured versus true volumes in irregularly shaped human coronary specimens. The measured volumes in the irregularly shaped human coronary specimens differed by 5.79% (median) from the true volumes (Figure 3). The error was higher than in phantoms (P ⫽ .036) and revealed a large variation (25th to 75th percentiles: ⫺0.01% to 9.23%). This was because of the difficulty of precise contour delineation of the inner and outer vessel border (often very irregular), which led to underestimation or overestimation. This was particularly problematic when acoustic shadowing from calcium deposits obscured the underlying wall.10,16 In these cases, the investigator could only speculate and would trace the border either too much inside (leading to underestimation) or outside the vessel (leading to overestimation). In addition, it was suggested that higher volume error variations in human coronary segments arteries are attributed to varying distances between the catheter and plaque.14 Similarly, Regar et al17 reported a more pronounced variability of measured neointimal areas (difference between the stent area and lumen area) in stents at 6-month follow-up. A median error less than 10% is probably acceptable for both clinical and research applications in interventional cardiology. For example, when compared with quantitative coronary angiography, visual estimates of lesion severity in coronary angiography were found to overestimate the severity of moderate stenosis, on average, by 30%.18 However, this volume error should be considered carefully, especially in arteriosclerotic progression/ regression studies, because the arteriosclerotic regression rate is quite small after long-term medical treatment.19,20 Previous studies have compared different volume measurement methods (IVUS vs histologic examination, IVUS vs angiography) without regard to the true volume of arteries, as in the current study. Matar et al21 measured the lumen volume of 13 human coronary vessels in vitro by 3D-IVUS and compared the values with the lumen volume derived from histologic section. The cross-sectional lumen area measured in the histologic sections was multiplied by vessel length. The group found a good correlation of 0.97. However, the absolute dimensions of histologic specimens may differ significantly
828 Sta¨hr et al
Journal of the American Society of Echocardiography August 2002
after tissue fixation22,23 and cannot be directly compared with IVUS measurements. Because of the inability to determine the true plaque or lumen volume clinically, interobserver and intraobserver variability testing8,24,25 and comparative studies with angiography26-29 were performed with 2-dimensional/3DIVUS measurements. Image Vessel Shape Versus True Shape The 3D-images of the curved phantoms were different from true shapes: The curve radius of the imaged phantoms was smaller than the curve radius of the true phantom (Figures 1 and 2). This is because the cross-sectional images were reconstructed on the basis of a fixed, straight axis (Figure 4). Theoretical 3D reconstructions for extreme catheter positions will make this clear: If, within a curvature, the catheter is precisely parallel to the vessel lumen, the longitudinal image will be presented as a straight segment (Figure 4, A). However, if the catheter is straightened within the same curvature (which, in a severe curve, is possible only when the catheter is not centered at the edges), the image will show the correct curve radius and the cross-sectional image will be displayed elliptically, especially in the proximal and distal parts of the segment (Figure 4, B). If the catheter bends in an intermediate curve between a straight line and the vessel’s curvature (which was the case in the current study), the displayed image curve radius is smaller than the true curve radius (Figure 4, C). Schwarzacher et al13 underestimated the true curve angle in curvatures of greater than 20 degrees. To perform spatially correct 3D reconstruction, and to overcome the insufficiency of conventional 3D-IVUS, the use of combined IVUS data and simultaneous biplane cinefluoroscopy30-32 has been suggested. Limitations
Figure 4 Longitudinal image reconstruction is dependent on catheter position within vessel lumen. A, If within a curvature, catheter is precisely parallel to vessel lumen, longitudinal image will be presented as straight segment. B, If catheter is straightened within same curvature (which, in severe curve, is possible only when catheter is not centered at edges), image will show correct curve radius, whereas cross-sectional image is displayed elliptically, especially in proximal and distal parts of segment. C, If catheter bends in an intermediate curve between straight line and vessel’s curvature (which was such in this study), displayed image curve radius is smaller than true curve radius.
Clinically, plaque volume measurements are required (defined as “combined media and intima without adventitia and connective tissue”). Because the pycnometer method was used, we measured the total volume of the coronary segments by IVUS, including intima, media, adventitia, and connective tissue volumes. Some tissue flaps at the outer vessel border may not have been included in the 3D IVUS volume analysis. We analyzed artery segments between 16.5 and 23 mm in length, which were removed from the atrioventricular groove or left anterior descending coronary artery groove. Thus, the segments may have lost their “true” bending. The volume error might have been larger when the IVUS volume measurements had been performed in the true (more severely curved) curvatures. Furthermore, there were no cardiac cycle pulsation simulations in either the phantoms or coronary
Journal of the American Society of Echocardiography Volume 15 Number 8
segments in our experimental setting.The image quality decreased after the digitization process with the current system compared with the original IVUS images. Even though this did not limit border detection, the current 3D system could benefit from an improved image quality. Investigations with larger catheters in smaller lumens, which was not addressed in this study, may add error to the volume measurements, as the catheter artifact in the image may hamper the lumen border detection. Commercially available catheters with higher frequencies (eg, 40 MHz) enhance blood speckle appearance and may interfere with accurate lumen border tracing.
CONCLUSIONS AND CLINICAL IMPLICATIONS With the current system, a median percentage deviation of 3D-IVUS-measured volumes from the true volumes in straight and curved phantoms and in coronary artery segments of less than 10% can be achieved. However, in the coronary specimens, this error shows a greater variance. We express our thanks to Peter Sta¨hr senior, MSME, and M. Lingner for their efforts to produce the phantoms. We also thank M. Rippin, PhD, Department of Statistics, Johannes Gutenberg-University, Mainz, Germany, for his statistical advice and Jorge Luna, MD, and Heidi Bonneau, RN, MS, for manuscript review.
REFERENCES 1. Holmes D, Fitzgerald P, Goldberg S, LaBlanche J, Lincoff AM, Savage M, et al. The PRESTO (Prevention of restenosis with tranilast and its outcomes) protocol: a double-blind, placebo-controlled trial. Am Heart J 2000;139:23-31. 2. Costa MA, Sabate M, Serrano P, van der Giessen WJ, Kozuma K, Kay IP, et al. The effect of 32P beta-radiotherapy on both vessel remodeling and neointimal hyperplasia after coronary balloon angioplasty and stenting: a three-dimensional intravascular ultrasound investigation. J Invasive Cardiol 2000;12: 113-20. 3. Sabate M, Serruys PW, van der Giessen WJ, Ligthart JM, Coen VL, Kay IP, et al. Geometric vascular remodeling after balloon angioplasty and beta- radiation therapy: a three-dimensional intravascular ultrasound study. Circulation 1999;100:1182-8. 4. Lu¨ scher TF, Zeiher A, Meinertz T, Hugenholtz PG, Quitzau KJ, Jenni R, et al. on behalf of the ENCORE trial investigators. J Cardiovasc Pharmacol 1997;30(suppl 3):48-52. 5. de Feyter PJ, Serruys PW, Davies MJ, Richardson P, Lubsen J, Oliver MF. Quantitative coronary angiography to measure progression and regression of coronary atherosclerosis: value, limitations, and implications for clinical trials. Circulation 1991;84:412-23. 6. Hong MK, Mintz GS, Popma JJ, Kent KM, Pichard AD, Satler LF, et al. Limitations of angiography for analyzing coronary atherosclerosis progression or regression. Ann Intern Med 1994;121:348-54.
Sta¨hr et al
829
7. Li W, von Birgelen C, di Mario C, Boersma E, Gussenhoven EJ, van der Putten N, et al. Semi-automatic contour detection for volumetric quantification of intracoronary ultrasound. In: Computers in Cardiology 1994. Los Angeles (CA): IEEE Computer Society Press; 1994:277-80. 8. von Birgelen C, de Vrey EA, Mintz GS, Nicosia A, Bruining N, Li W, et al. ECG-gated three-dimensional intravascular ultrasound: feasibility and reproducibility of the automated analysis of coronary lumen and atherosclerotic plaque dimensions in humans. Circulation 1997;96:2944-52. 9. von Birgelen C, Di Mario C, Li W, Schuurbiers JC, Slager CJ, de Feyter PJ, et al. Morphometric analysis in three-dimensional intracoronary ultrasound: an in vitro and in vivo study performed with a novel system for the contour detection of lumen and plaque. Am Heart J 1996;132:516-27. 10. Roelandt JR, Di Mario C, Pandian NG, Wenguang L, Keane D, Slager CJ, et al. Three-dimensional reconstruction of intracoronary ultrasound images: rationale, approaches, problems, and directions. Circulation 1994;90:1044-55. 11. Krass S, Brennecke R, Voigtla¨nder T, Sta¨hr P, Rupprecht H, Meyer J. Assessment of image quality of intracoronary ultrasound systems with tissue-equivalent vessel phantoms. Los Angeles (CA): IEEE Computer Society Press; 1994. 12. Sta¨hr P, Rupprecht HJ, Voigtla¨nder T, Kearney P, Erbel R, Koch L, et al. Importance of calibration for diameter and area determination by intravascular ultrasound. Int J Card Imaging 1996;12:221-9. 13. Schwarzacher SP, Honda Y, Metz JA, Asvar CA, Fitzgerald PJ, Yock PG. Impact of curve distortion errors on intravascular ultrasound measurements and three-dimensional reconstructions. Am J Cardiol 1997;79:384-7. 14. Schuurbiers JC, von Birgelen C, Wentzel JJ, Bom N, Serruys PW, de Feyter PJ, et al. On the IVUS plaque volume error in coronary arteries when neglecting curvature. Ultrasound Med Biol 2000;26:1403-11. 15. Wiet SP, Vonesh MJ, Waligora MJ, Kane BJ, McPherson DD. The effect of vascular curvature on three-dimensional reconstruction of intravascular ultrasound images. Ann Biomed Eng 1996;24:695-701. 16. Mintz GS, Popma JJ, Pichard AD, Kent KM, Satler LF, Chuang YC, et al. Limitations of angiography in the assessment of plaque distribution in coronary artery disease: a systematic study of target lesion eccentricity in 1446 lesions. Circulation 1996;93:924-31. 17. Regar E, Werner F, Siebert U, Rieber J, Theisen K, Mudra H, et al. Reproducibility of neointima quantification with motorized intravascular ultrasound pullback in stented coronary arteries. Am Heart J 2000;139:632-7. 18. Fleming RM, Kirkeeide RL, Smalling RW, Gould KL. Patterns in visual interpretation of coronary arteriograms as detected by quantitative coronary arteriography. J Am Coll Cardiol 1991;18:945-51. 19. Jukema JW, Bruschke AV, van Boven AJ, Reiber JH, Bal ET, Zwinderman AH, et al. Effects of lipid lowering by pravastatin on progression and regression of coronary artery disease in symptomatic men with normal to moderately elevated serum cholesterol levels: the Regression Growth Evaluation Statin Study (REGRESS). Circulation 1995;91:2528-40. 20. Gotto AM. Lipid lowering, regression, and coronary events: a review of the Interdisciplinary Council on Lipids and Cardiovascular Risk Intervention, Seventh Council meeting. Circulation 1995;92:646-56. 21. Matar FA, Mintz GS, Douek P, Farb A, Virmani R, Javier SP, et al. Coronary artery lumen volume measurement using
830 Sta¨hr et al
22.
23.
24.
25.
26.
three-dimensional intravascular ultrasound: validation of a new technique [see comments]. Cathet Cardiovasc Diagn 1994;33:214-20. Siegel RJ, Swan K, Edwalds G, Fishbein MC. Limitations of postmortem assessment of human coronary artery size and luminal narrowing: differential effects of tissue fixation and processing on vessels with different degrees of atherosclerosis. J Am Coll Cardiol 1985;5:342-6. Sta¨hr P, Rupprecht HJ, Voigtla¨nder T, Otto M, Rudigier K, Erbel R, et al. Comparison of normal and diseased pulmonary artery morphology by intravascular ultrasound and histological examination. Int J Card Imaging 1999;15:221-31. Hausmann D, Lundkvist AJ, Friedrich GJ, Mullen WL, Fitzgerald PJ, Yock PG. Intracoronary ultrasound imaging: intraobserver and interobserver variability of morphometric measurements. Am Heart J 1994;128:674-80. Hagenaars T, Gussenhoven EJ, van Essen JA, Seelen J, Honkoop J, van der Lugt A. Reproducibility of volumetric quantification in intravascular ultrasound images. Ultrasound Med Biol 2000;26:367-74. von Birgelen C, Kutryk MJ, Gil R, Ozaki Y, Di Mario C, Roelandt JR, et al. Quantification of the minimal luminal cross-sectional area after coronary stenting by two- and threedimensional intravascular ultrasound versus edge detection and videodensitometry. Am J Cardiol 1996;78:520-5.
Journal of the American Society of Echocardiography August 2002
27. Dussaillant GR, Mintz GS, Pichard AD, Kent KM, Satler LF, Popma JJ, et al. Small stent size and intimal hyperplasia contribute to restenosis: a volumetric intravascular ultrasound analysis. J Am Coll Cardiol 1995;26:720-4. 28. Dhawale PJ, Wilson DL, Hodgson JM. Optimal data acquisition for volumetric intracoronary ultrasound. Cathet Cardiovasc Diagn 1994;32:288-99. 29. Hausmann D, Friedrich G, Sudhir K, Mullen WL, Soni B, Fitzgerald PJ, et al. Three dimensional intravascular ultrasound imaging with automated border detection using 2.9 F catheters. J Am Coll Cardiol 1994:174A. 30. Slager CJ, Wentzel JJ, Schuurbiers JC, Oomen JA, Kloet J, Krams R, et al. True 3-dimensional reconstruction of coronary arteries in patients by fusion of angiography and IVUS (ANGUS) and its quantitative validation. Circulation 2000;102: 511-6. 31. Evans JL, Ng KH, Wiet SG, Vonesh MJ, Burns WB, Radvany MG, et al. Accurate three-dimensional reconstruction of intravascular ultrasound data: spatially correct three-dimensional reconstructions. Circulation 1996;93:567-76. 32. Wahle A, Prause PM, DeJong SC, Sonka M. Geometrically correct 3-D reconstruction of intravascular ultrasound images by fusion with biplane angiography—methods and validation. IEEE Trans Med Imaging 1999;18:686-99.
Background: Three-dimensional intravascular ultrasound (IVUS) is used for volumetric assessment of arteriosclerotic plaque burden and restenotic tissue at follow-up after coronary interventions. However, the accuracy of these measurements, especially in tortuous vessels, is unclear. Methods: A commercially available electrocardiogram (ECG)-gated 3-dimensional-IVUS system was tested in volume-validated straight and curved hydrocolloid phantoms and in volume-validated coronary specimens. Catheter withdrawal (30 MHz, 3.2F) was triggered using standardized ECG source with 0.2-mm step intervals per cardiac cycle simulation.
A
t present, 3-dimensional (3D)-intravascular ultrasound (IVUS) is used as an essential research tool in the evaluation of specific antirestenosis therapies.1-3 It may also serve as a valuable tool in the understanding and assessment of the regression/progression of arteriosclerosis over time in a given patient population. For example, a 3D-IVUS investigation has been initiated to evaluate the response of arteriosclerosis to treatment with cerivastatin sodium, nifedipine, or both.4 In contrast, quantitative coronary angiography has known limitations in demonstrating small differences in vessel architecture, as seen in secondary prevention trials.5,6 In addition, semiautomated 3D-IVUS systems with electrocardiogram (ECG) triggering for the detection of the intimal leading edge and the external From the 2nd Medical Clinic (Cardiology) and Department of Pathology (M.O.), Johannes Gutenberg-University, Mainz, Germany; and Stanford University Medical School (P.S., P.J.F.), California. Supported by a Feodor-Lynen grant from the Alexander von Humboldt-Stiftung (P.S.). Reprint requests: Peter J. Fitzgerald, MD, PhD, Stanford University School of Medicine, Division of Cardiovascular Medicine, 300 Pasteur Dr, Room H3554, Stanford, CA 94305 (E-mail: [email protected]). Copyright 2002 by the American Society of Echocardiography. 0894-7317/2002/$35.00 ⫹ 0 27/1/120700 doi:10.1067/mje.2002.120700
Results: On the basis of automated phantom volume measurements, IVUS overestimated true phantom volume (relative error ⴝ [measured V ⴚ true V]/true V ⴛ 100) by a median of 0.9%, 0.25%, and 1.96% for straight, mildly curved, and severely curved segments, respectively. The true volume of the coronary specimens was overestimated by a median of 5.79%. Conclusion: A median percentage deviation of 3-dimensional-IVUS-measured volumes from the true volumes of less than 10% in phantoms and coronary artery segments can be achieved. (J Am Soc Echocardiogr 2002;15:823-30.)
vessel contour (media-adventitia boundary) have been developed for volume quantitation.7-9 However, with all these volume measurement techniques, curvatures may influence the IVUS-quantitation of vessel lumen and plaque volume.10 The aim of this study was to identify volume variations in ECG-gated 3D-IVUS-measurements in both straight and curved hydrocolloid phantoms and in irregularly shaped coronary artery specimens. These validations may contribute to the accuracy of future research studies investigating volumetric changes in arteriosclerotic plaque over time and the extent of restenotic tissue proliferation.
METHODS IVUS Imaging and ECG-gated 3D-IVUS Image Acquisition and Analysis Imaging of the phantoms and coronary artery segments was performed in vitro using a routinely available mechanical IVUS system (ClearView cardiovascular imaging system, Boston Scientific Corp, Maple Grove, Minn) with a 30 MHz catheter (3.2F Ultracross and Scimed, Boston Scientific Corp, Maple Grove, Minn). The catheter used was sheath based, meaning that the transducer element could be moved in a transparent trajectory.
823
Journal of the American Society of Echocardiography August 2002
824 Sta¨hr et al
Figure 1 Longitudinally reconstructed intravascular ultrasound images of 3 different phantom types (A, straight; B, mildly curved; C, severely curved) and D, coronary specimen. Note almost straight shape of curved phantom images (B, C) and catheter touching wall. Table 1 Absolute values of phantom dimensions (4 phantoms of each type with 10 IVUS measurements in each phantom)
Phantom type (n ⴝ 4 each)
Straight P Mildly CP Severely CP
True lumen diameter (mm)
True central lumen length (mm)
True volume (mean ⴞ SD) (mm3)
Measured volume (mean ⴞ SD) (mm3)
Absolute difference (mean ⴞ SD) (mm3)
2.7 2.7 2.7
20 20 20
279.02 ⫾ 2.01 257.20 ⫾ 2.17 264.67 ⫾ 3.52
280.52 ⫾ 8.85 257.55 ⫾ 11.1 270.64 ⫾ 1.44
1.5 ⫾ 8.24 0.36 ⫾ 10.99 5.97 ⫾ 9.64
IVUS, Intravascular ultrasound; C, curved; P, phantom.
The ECG-gated 3D-IVUS image acquisition and digitization (TomTec Imaging Systems Inc, Unterschleissheim, Germany) has been described in detail elsewhere.8 The transducer, when positioned in the phantom or coronary artery segment, was withdrawn by a pullback device in step intervals of 0.2 mm per cardiac cycle, triggered by a standardized ECG source. IVUS images were acquired and digitized 40 ms after the peak of the R-wave. Consecutively, the transducer was withdrawn to acquire the next image. Each set of digitized IVUS images was analyzed off-line with an automated, computerized contour detection algorithm.7,8 Two longitudinal sections were constructed and contours corresponding to the inner lumen border and outer phantom/segment border were automatically identified by a minimal cost algorithm7,8 (Figure 1). If necessary, these longitudinal contours were edited with computer assistance. The longitudinal contours were transformed to individual edge points on the cross-sectional images. On the basis of these edge points, the system automatically suggested the border contour in each cross-sectional image (Figure 1). In cases with incomplete border evaluation, the
contour presented could be manually edited by the operator. The wall cross-sectional area (A ⫽ outer cross-sectional area ⫺ lumen cross-sectional area) was calculated for each planar image. Volume of the phantom/segment was calculated by: V⫽n⌺
i⫽1
(A1 ⫻ H)
where H is the thickness of the slice (represented by a single tomographic IVUS image) and n is the number of IVUS images in the 3D data set.8 Phantom Volume Measurements by IVUS Phantoms used for volume measurements were made from a hydrocolloid (Optiloid, Riss-Dental, Hanau, Germany) with tissue-like ultrasound properties.11 All phantoms had a central lumen length of 20 mm, an inner lumen diameter of 2.7 mm, and an outer diameter of 5 mm (Table 1). Three different types of phantoms were produced: straight, mildly curved (radius 30 mm), and severely curved (radius 10 mm) (Figure 2).
Journal of the American Society of Echocardiography Volume 15 Number 8
Sta¨hr et al
825
Figure 2 Sketch of true dimensions of different phantom types: A, straight; B, mildly curved; C, severely curved.
Within the 3 types, 4 phantoms each were used with 10 pullbacks being performed for each phantom. Volumes were measured off-line using the automated, computerized system described previously. Coronary Artery Segment Volume Measurements by IVUS Ten straight segments of coronary arteries (5 left anterior descending coronary artery, 2 left circumflex coronary artery, 3 right coronary artery) of 5 patients obtained by autopsy were investigated by IVUS after they were removed from the atrioventricular groove or left anterior descending coronary artery groove. Ten pullbacks in each segment were performed. Five of the 10 segment contours derived automatically were optimized manually in the planar images by one of the authors (P.S.). Thus, volume measurements from automated and manually corrected border detection could be compared. Quantitative volume measurements with manual correction were repeated at least 6 weeks apart and used to assess their reliability. The correlation coefficient obtained from linear regression analysis and the percent error obtained by taking the absolute difference divided by the initial measurements were used to express the intraobserver variability. The intraobserver correlation coefficient and percent error were r2 ⫽ 0.99 and 1.2% ⫾ 0.8% (mean ⫾ SD). The sheath-based catheter was centered, without the use of a guide wire, at the inlet and outlet of the phantoms or artery segments. The pullbacks were performed in 20°C water. After each pullback, it was mandatory to supply a reference distance for the 3D software. To provide reliable referencing, the diameter of a hydrocolloid phantom (5 mm) was used for distance calibration. This reference was included at the end of each pullback in both phantoms and artery segments. As a result of this procedure, differences in sound velocities of blood and water, leading to errors in terms of overestimation or underestimation, could be minimized.12
Validation of Phantom and Coronary Segment Volumes The true volume of the phantoms and artery segments was determined by pycnometer measurements. The method is on the basis of weighing (Microgram weighing machine, Sartorius, Goettingen, Germany) the water volume displaced by the body after immersion into the definite volume of the water-filled pycnometer (20°C). The water weight divided by the water density yields the body volume. The accuracy of measurements is within a range of ⫾0.07%. Statistical Analysis Results are given as median with the 25th, 75th, 10th, and 90th percentiles or as mean ⫾ SD. Systematic differences of the relative volume errors between the different kinds of phantoms or arteries were tested using the t test for unpaired observations. When normality or equal variance testing failed, the Mann-Whitney rank sum test was applied. A P value ⬍ .05 was considered to be statistically significant.
RESULTS Phantom Volume Measurements by IVUS Sample images from each phantom type are shown in Figure 2. The true phantom curves are not visible in the longitudinal IVUS images; the curved phantoms almost appear straight. However, because of the curvatures, the catheter is forced toward the wall (Figure 1, B and C). The automatic contour detection matched the phantom borders very well, so that manual contour optimization was not necessary. The absolute true volumes (evaluated by pycnometer measurements) and volumes (measured by IVUS) are listed in Table
Journal of the American Society of Echocardiography August 2002
826 Sta¨hr et al
Figure 3 Three-dimensional–intravascular ultrasound volume measurements: Deviation of measured volume from true phantom or coronary artery segment volume as expressed by relative error (%) ⫽ (measured V ⫺ true V)/true V ⫻ 100. V, Volume. Table 2 Absolute values (after manual contour optimization) of coronary artery specimens (with 5-10 IVUS pullbacks in each segment) Artery no.
Mean measured LD (mm)
Mean measured central LL (mm)
True V (mm3)
Measured V (mean ⴞ SD) (mm3)
Absolute difference (mean ⴞ SD) (mm3)
1 2 3 4 5 6 7 8 9 10
2 2.6 1.5 2.9 2.3 2.5 2.2 1.8 2.3 2.5
19.8 20.8 17.6 22.2 20.2 18 16.5 17.8 19 23
289.35 301.29 137.76 337.61 190.34 210.38 86.16 146.00 258.47 282.83
284.84 ⫾ 8.87 260.69 ⫾ 14.10 146.56 ⫾ 12.43 331.75 ⫾ 8.84 204.05 ⫾ 7.62 221.91 ⫾ 3.98 101.64 ⫾ 3.67 160.03 ⫾ 4.39 270.61 ⫾ 1.58 304.30 ⫾ 2.41
⫺4.51 ⫾ 8.87 ⫺40.60 ⫾ 14.10 8.80 ⫾ 12.43 ⫺5.86 ⫾ 8.84 13.71 ⫾ 7.62 11.53 ⫾ 3.98 15.48 ⫾ 3.67 14.03 ⫾ 4.39 12.14 ⫾ 1.58 21.47 ⫾ 2.41
IVUS, Intravascular ultrasound; LD, lumen diameter; LL, lumen length; V, volume.
1. On the basis of the automated volume measurements, 3D-IVUS overestimated the true volume (relative error ⫽ [measured ⫺ true]/true volume ⫻ 100) by 0.9% (median, 25th to 75th percentiles: ⫺1.15% to 2.45%) in the straight phantoms and by 0.25% (25th to 75th percentiles: ⫺2.21% to 2.6%) in the mildly curved phantoms (P ⫽ .65, Figure 3). In the severely curved segments, the true volume was overestimated by 1.96% (25th to 75th percentiles: ⫺1.15% to 5.27%), which is significantly different from the relative error in both the straight and mildly curved phantoms (P ⫽ .035 and P ⫽ .045, respectively, Figure 3).
Coronary Artery Segment Volume Measurements by IVUS Table 2 lists dimensions and volumes of the artery segments measured by IVUS (after manual optimizing of contour delineation). IVUS measurements overestimated true segment volumes, evaluated by pycnometer measurements, by a median relative error of 5.79% (25th to 75th percentiles: ⫺0.01% to 9.23%) (Figure 3). Compared with the error observed in the phantom measurements, the relative error in the coronary artery segments demonstrated a wider variability (wider 10th, 25th, 75th, and 90th percentiles, Figure 3).
Journal of the American Society of Echocardiography Volume 15 Number 8
Sta¨hr et al
827
IVUS measurements on the basis of automatic contour delineation alone, without manual optimization, showed a higher relative error (median) of 8.1% (25th to 75th percentiles: 3.37% to 14.63%), P ⫽ .018.
sectional area in curved phantoms, reported an overestimation of the true cross-sectional area by up to 18% in higher curvatures. The differences in errors may be partly because of different distances between the transducer and the vessel wall or plaque.14
DISCUSSION
Volume Measurements in Coronary Artery Segments
Phantom Volume Measurements 3D-IVUS overestimated the true volume of the straight and mildly curved phantoms (curve radius 30 mm) with a median of up to 0.9% (Figure 3) and not higher than 1.96% in the severely curved segments (curve radius 10 mm). The higher overestimation in severely curved segments was because of geometric factors: compression/expansion of the vessel and oblique cross-sectional areas within the curvature.10,13 Similar to our results, luminal volume measured in a straight paraffin tubular phantom by von Birgelen et al,9 using the same 3D-system without ECG gating, was overestimated by 0.25% to 1.72%. It should be stressed that von Birgelen’s investigation was aimed at the lumen volume in contrast to the total phantom volume in our investigation. For measuring the phantom volume, a contour detection of the inner and outer border is necessary, with the potential of increased systematic error. In this study, we used a sheath-based catheter with the catheter moved inside a transparent trajectory. The transducer traverses along a path that is relatively straight compared with the local vessel curvature.14 According to Schuurbiers et al,14 the volume error with this system was primarily dependent on the curvature of the catheter pullback trajectory; vessel tortuosity itself played a role only in so far as it affected the local curvature of the pullback trajectory. Furthermore, with increasing curvature of the pullback trajectory, the volume error was increasing, which is in line with our study. Hence, the error in the same curve may be even higher when using a nonsheath-based catheter system in which the catheter is moved as a whole in the vessel. Other groups have described an overestimation of the lumen volume or cross-sectional area of curved segments: Wiet et al15 found an overestimation of the true lumen volume in curved tubes (length 25 mm) of 1% (for a curve radius of 100 mm) and 35% (for a curve radius of 20 mm) measured with another 3D-IVUS system. The group recommended only lower curvatures (curve radius ⬎ 60 mm) for 3D-IVUS-measurements. In our study, even in segments with severe curvature (radius of 10 mm), we found an overestimation of less than 10%. In addition, Schwarzacher et al,13 measuring the cross-
In addition to phantoms, we examined the variations of 3D-IVUS measured versus true volumes in irregularly shaped human coronary specimens. The measured volumes in the irregularly shaped human coronary specimens differed by 5.79% (median) from the true volumes (Figure 3). The error was higher than in phantoms (P ⫽ .036) and revealed a large variation (25th to 75th percentiles: ⫺0.01% to 9.23%). This was because of the difficulty of precise contour delineation of the inner and outer vessel border (often very irregular), which led to underestimation or overestimation. This was particularly problematic when acoustic shadowing from calcium deposits obscured the underlying wall.10,16 In these cases, the investigator could only speculate and would trace the border either too much inside (leading to underestimation) or outside the vessel (leading to overestimation). In addition, it was suggested that higher volume error variations in human coronary segments arteries are attributed to varying distances between the catheter and plaque.14 Similarly, Regar et al17 reported a more pronounced variability of measured neointimal areas (difference between the stent area and lumen area) in stents at 6-month follow-up. A median error less than 10% is probably acceptable for both clinical and research applications in interventional cardiology. For example, when compared with quantitative coronary angiography, visual estimates of lesion severity in coronary angiography were found to overestimate the severity of moderate stenosis, on average, by 30%.18 However, this volume error should be considered carefully, especially in arteriosclerotic progression/ regression studies, because the arteriosclerotic regression rate is quite small after long-term medical treatment.19,20 Previous studies have compared different volume measurement methods (IVUS vs histologic examination, IVUS vs angiography) without regard to the true volume of arteries, as in the current study. Matar et al21 measured the lumen volume of 13 human coronary vessels in vitro by 3D-IVUS and compared the values with the lumen volume derived from histologic section. The cross-sectional lumen area measured in the histologic sections was multiplied by vessel length. The group found a good correlation of 0.97. However, the absolute dimensions of histologic specimens may differ significantly
828 Sta¨hr et al
Journal of the American Society of Echocardiography August 2002
after tissue fixation22,23 and cannot be directly compared with IVUS measurements. Because of the inability to determine the true plaque or lumen volume clinically, interobserver and intraobserver variability testing8,24,25 and comparative studies with angiography26-29 were performed with 2-dimensional/3DIVUS measurements. Image Vessel Shape Versus True Shape The 3D-images of the curved phantoms were different from true shapes: The curve radius of the imaged phantoms was smaller than the curve radius of the true phantom (Figures 1 and 2). This is because the cross-sectional images were reconstructed on the basis of a fixed, straight axis (Figure 4). Theoretical 3D reconstructions for extreme catheter positions will make this clear: If, within a curvature, the catheter is precisely parallel to the vessel lumen, the longitudinal image will be presented as a straight segment (Figure 4, A). However, if the catheter is straightened within the same curvature (which, in a severe curve, is possible only when the catheter is not centered at the edges), the image will show the correct curve radius and the cross-sectional image will be displayed elliptically, especially in the proximal and distal parts of the segment (Figure 4, B). If the catheter bends in an intermediate curve between a straight line and the vessel’s curvature (which was the case in the current study), the displayed image curve radius is smaller than the true curve radius (Figure 4, C). Schwarzacher et al13 underestimated the true curve angle in curvatures of greater than 20 degrees. To perform spatially correct 3D reconstruction, and to overcome the insufficiency of conventional 3D-IVUS, the use of combined IVUS data and simultaneous biplane cinefluoroscopy30-32 has been suggested. Limitations
Figure 4 Longitudinal image reconstruction is dependent on catheter position within vessel lumen. A, If within a curvature, catheter is precisely parallel to vessel lumen, longitudinal image will be presented as straight segment. B, If catheter is straightened within same curvature (which, in severe curve, is possible only when catheter is not centered at edges), image will show correct curve radius, whereas cross-sectional image is displayed elliptically, especially in proximal and distal parts of segment. C, If catheter bends in an intermediate curve between straight line and vessel’s curvature (which was such in this study), displayed image curve radius is smaller than true curve radius.
Clinically, plaque volume measurements are required (defined as “combined media and intima without adventitia and connective tissue”). Because the pycnometer method was used, we measured the total volume of the coronary segments by IVUS, including intima, media, adventitia, and connective tissue volumes. Some tissue flaps at the outer vessel border may not have been included in the 3D IVUS volume analysis. We analyzed artery segments between 16.5 and 23 mm in length, which were removed from the atrioventricular groove or left anterior descending coronary artery groove. Thus, the segments may have lost their “true” bending. The volume error might have been larger when the IVUS volume measurements had been performed in the true (more severely curved) curvatures. Furthermore, there were no cardiac cycle pulsation simulations in either the phantoms or coronary
Journal of the American Society of Echocardiography Volume 15 Number 8
segments in our experimental setting.The image quality decreased after the digitization process with the current system compared with the original IVUS images. Even though this did not limit border detection, the current 3D system could benefit from an improved image quality. Investigations with larger catheters in smaller lumens, which was not addressed in this study, may add error to the volume measurements, as the catheter artifact in the image may hamper the lumen border detection. Commercially available catheters with higher frequencies (eg, 40 MHz) enhance blood speckle appearance and may interfere with accurate lumen border tracing.
CONCLUSIONS AND CLINICAL IMPLICATIONS With the current system, a median percentage deviation of 3D-IVUS-measured volumes from the true volumes in straight and curved phantoms and in coronary artery segments of less than 10% can be achieved. However, in the coronary specimens, this error shows a greater variance. We express our thanks to Peter Sta¨hr senior, MSME, and M. Lingner for their efforts to produce the phantoms. We also thank M. Rippin, PhD, Department of Statistics, Johannes Gutenberg-University, Mainz, Germany, for his statistical advice and Jorge Luna, MD, and Heidi Bonneau, RN, MS, for manuscript review.
REFERENCES 1. Holmes D, Fitzgerald P, Goldberg S, LaBlanche J, Lincoff AM, Savage M, et al. The PRESTO (Prevention of restenosis with tranilast and its outcomes) protocol: a double-blind, placebo-controlled trial. Am Heart J 2000;139:23-31. 2. Costa MA, Sabate M, Serrano P, van der Giessen WJ, Kozuma K, Kay IP, et al. The effect of 32P beta-radiotherapy on both vessel remodeling and neointimal hyperplasia after coronary balloon angioplasty and stenting: a three-dimensional intravascular ultrasound investigation. J Invasive Cardiol 2000;12: 113-20. 3. Sabate M, Serruys PW, van der Giessen WJ, Ligthart JM, Coen VL, Kay IP, et al. Geometric vascular remodeling after balloon angioplasty and beta- radiation therapy: a three-dimensional intravascular ultrasound study. Circulation 1999;100:1182-8. 4. Lu¨ scher TF, Zeiher A, Meinertz T, Hugenholtz PG, Quitzau KJ, Jenni R, et al. on behalf of the ENCORE trial investigators. J Cardiovasc Pharmacol 1997;30(suppl 3):48-52. 5. de Feyter PJ, Serruys PW, Davies MJ, Richardson P, Lubsen J, Oliver MF. Quantitative coronary angiography to measure progression and regression of coronary atherosclerosis: value, limitations, and implications for clinical trials. Circulation 1991;84:412-23. 6. Hong MK, Mintz GS, Popma JJ, Kent KM, Pichard AD, Satler LF, et al. Limitations of angiography for analyzing coronary atherosclerosis progression or regression. Ann Intern Med 1994;121:348-54.
Sta¨hr et al
829
7. Li W, von Birgelen C, di Mario C, Boersma E, Gussenhoven EJ, van der Putten N, et al. Semi-automatic contour detection for volumetric quantification of intracoronary ultrasound. In: Computers in Cardiology 1994. Los Angeles (CA): IEEE Computer Society Press; 1994:277-80. 8. von Birgelen C, de Vrey EA, Mintz GS, Nicosia A, Bruining N, Li W, et al. ECG-gated three-dimensional intravascular ultrasound: feasibility and reproducibility of the automated analysis of coronary lumen and atherosclerotic plaque dimensions in humans. Circulation 1997;96:2944-52. 9. von Birgelen C, Di Mario C, Li W, Schuurbiers JC, Slager CJ, de Feyter PJ, et al. Morphometric analysis in three-dimensional intracoronary ultrasound: an in vitro and in vivo study performed with a novel system for the contour detection of lumen and plaque. Am Heart J 1996;132:516-27. 10. Roelandt JR, Di Mario C, Pandian NG, Wenguang L, Keane D, Slager CJ, et al. Three-dimensional reconstruction of intracoronary ultrasound images: rationale, approaches, problems, and directions. Circulation 1994;90:1044-55. 11. Krass S, Brennecke R, Voigtla¨nder T, Sta¨hr P, Rupprecht H, Meyer J. Assessment of image quality of intracoronary ultrasound systems with tissue-equivalent vessel phantoms. Los Angeles (CA): IEEE Computer Society Press; 1994. 12. Sta¨hr P, Rupprecht HJ, Voigtla¨nder T, Kearney P, Erbel R, Koch L, et al. Importance of calibration for diameter and area determination by intravascular ultrasound. Int J Card Imaging 1996;12:221-9. 13. Schwarzacher SP, Honda Y, Metz JA, Asvar CA, Fitzgerald PJ, Yock PG. Impact of curve distortion errors on intravascular ultrasound measurements and three-dimensional reconstructions. Am J Cardiol 1997;79:384-7. 14. Schuurbiers JC, von Birgelen C, Wentzel JJ, Bom N, Serruys PW, de Feyter PJ, et al. On the IVUS plaque volume error in coronary arteries when neglecting curvature. Ultrasound Med Biol 2000;26:1403-11. 15. Wiet SP, Vonesh MJ, Waligora MJ, Kane BJ, McPherson DD. The effect of vascular curvature on three-dimensional reconstruction of intravascular ultrasound images. Ann Biomed Eng 1996;24:695-701. 16. Mintz GS, Popma JJ, Pichard AD, Kent KM, Satler LF, Chuang YC, et al. Limitations of angiography in the assessment of plaque distribution in coronary artery disease: a systematic study of target lesion eccentricity in 1446 lesions. Circulation 1996;93:924-31. 17. Regar E, Werner F, Siebert U, Rieber J, Theisen K, Mudra H, et al. Reproducibility of neointima quantification with motorized intravascular ultrasound pullback in stented coronary arteries. Am Heart J 2000;139:632-7. 18. Fleming RM, Kirkeeide RL, Smalling RW, Gould KL. Patterns in visual interpretation of coronary arteriograms as detected by quantitative coronary arteriography. J Am Coll Cardiol 1991;18:945-51. 19. Jukema JW, Bruschke AV, van Boven AJ, Reiber JH, Bal ET, Zwinderman AH, et al. Effects of lipid lowering by pravastatin on progression and regression of coronary artery disease in symptomatic men with normal to moderately elevated serum cholesterol levels: the Regression Growth Evaluation Statin Study (REGRESS). Circulation 1995;91:2528-40. 20. Gotto AM. Lipid lowering, regression, and coronary events: a review of the Interdisciplinary Council on Lipids and Cardiovascular Risk Intervention, Seventh Council meeting. Circulation 1995;92:646-56. 21. Matar FA, Mintz GS, Douek P, Farb A, Virmani R, Javier SP, et al. Coronary artery lumen volume measurement using
830 Sta¨hr et al
22.
23.
24.
25.
26.
three-dimensional intravascular ultrasound: validation of a new technique [see comments]. Cathet Cardiovasc Diagn 1994;33:214-20. Siegel RJ, Swan K, Edwalds G, Fishbein MC. Limitations of postmortem assessment of human coronary artery size and luminal narrowing: differential effects of tissue fixation and processing on vessels with different degrees of atherosclerosis. J Am Coll Cardiol 1985;5:342-6. Sta¨hr P, Rupprecht HJ, Voigtla¨nder T, Otto M, Rudigier K, Erbel R, et al. Comparison of normal and diseased pulmonary artery morphology by intravascular ultrasound and histological examination. Int J Card Imaging 1999;15:221-31. Hausmann D, Lundkvist AJ, Friedrich GJ, Mullen WL, Fitzgerald PJ, Yock PG. Intracoronary ultrasound imaging: intraobserver and interobserver variability of morphometric measurements. Am Heart J 1994;128:674-80. Hagenaars T, Gussenhoven EJ, van Essen JA, Seelen J, Honkoop J, van der Lugt A. Reproducibility of volumetric quantification in intravascular ultrasound images. Ultrasound Med Biol 2000;26:367-74. von Birgelen C, Kutryk MJ, Gil R, Ozaki Y, Di Mario C, Roelandt JR, et al. Quantification of the minimal luminal cross-sectional area after coronary stenting by two- and threedimensional intravascular ultrasound versus edge detection and videodensitometry. Am J Cardiol 1996;78:520-5.
Journal of the American Society of Echocardiography August 2002
27. Dussaillant GR, Mintz GS, Pichard AD, Kent KM, Satler LF, Popma JJ, et al. Small stent size and intimal hyperplasia contribute to restenosis: a volumetric intravascular ultrasound analysis. J Am Coll Cardiol 1995;26:720-4. 28. Dhawale PJ, Wilson DL, Hodgson JM. Optimal data acquisition for volumetric intracoronary ultrasound. Cathet Cardiovasc Diagn 1994;32:288-99. 29. Hausmann D, Friedrich G, Sudhir K, Mullen WL, Soni B, Fitzgerald PJ, et al. Three dimensional intravascular ultrasound imaging with automated border detection using 2.9 F catheters. J Am Coll Cardiol 1994:174A. 30. Slager CJ, Wentzel JJ, Schuurbiers JC, Oomen JA, Kloet J, Krams R, et al. True 3-dimensional reconstruction of coronary arteries in patients by fusion of angiography and IVUS (ANGUS) and its quantitative validation. Circulation 2000;102: 511-6. 31. Evans JL, Ng KH, Wiet SG, Vonesh MJ, Burns WB, Radvany MG, et al. Accurate three-dimensional reconstruction of intravascular ultrasound data: spatially correct three-dimensional reconstructions. Circulation 1996;93:567-76. 32. Wahle A, Prause PM, DeJong SC, Sonka M. Geometrically correct 3-D reconstruction of intravascular ultrasound images by fusion with biplane angiography—methods and validation. IEEE Trans Med Imaging 1999;18:686-99.