ABSTRACT: Objective. To evaluate the flow characteristics and accuracy for the detection of patent or occluded coronary artery bypass grafts (CABG) with multi-slice flow study of electron-beam tomography (EBT). Methods. One hundred and twenty-three patients who had undergone CABG surgery were enrolled in this study. Flow datasets were assessed with time-density curves by EBT. The EBT results were blindly compared with post-operative cardiac catheterizations in 26 patients. Results. Image quality was adequate to evaluate in 111 patients (90.2%). Flow curves of bypass grafts were technically adequate in 265 of 309 (85.8%) saphenous-vein grafts (SVG) and 35 of 56 (62.5%) internal mammary artery (IMA) grafts (p < 0.05). In comparison to conventional angiographic results, EBT correctly identified 14 of 16 occluded grafts (sensitivity, 87.5%) and 68 of 75 patent grafts (specificity, 90.7%), yielding an accuracy of 90.1%. The intra-graft flows of the IMA and SVG were 4.9 ± 2.2 ml/min/g and 6.9 ± 2.8 ml/min/g, respectively (p < 0.001), which was 31.6 ± 20.4% and 39.4 ± 21.9% of the ascending aorta’s flow (16.7 ± 5.0 ml/min/g) (p < 0.001). Conclusion. EBT flow study can be used in the assessment of CABG patency and quantification of intra-graft flow of patent CABG vessels.

       Electron-beam angiography (EBA) with three-dimensional (3D) reconstruction is a new, noninvasive cardiac imaging modality that has been used to determine patency of coronary artery bypass grafts (CABG).1–3 EBA permits noninvasive identification of the entire course and anastomoses of CABG anatomy and determination of bypass graft occlusion or relevant stenosis with high accuracy; thus, EBA may be the first clinically practical noninvasive alternative to conventional coronary arteriography for evaluating the patency of the CABG.1 Another promising protocol for electron-beam tomography (EBT), which has been used to assess CABG patency, is multi-slice flow studies. Previous reports found that the sensitivity of EBT flow study was 93–96% and the specificity was 89–97%.4,5 The criteria for graft patency on EBA study were that the CABG could be identified and reconstructed three-dimensionally,1–3 and that contrast opacification was found on at least two non-contiguous levels. The criteria for graft patency on EBT flow study were that the graft was visualized at two or more levels in the distribution of the vessel bypassed or that its time-density curve (TDC) was morphologically similar to that of the ascending aorta.4,5
       Unfortunately, none of these reported data quantified the flow characteristics of patent CABG vessels. This study was designed to evaluate the accuracy and interobserver agreement of EBT flow study, and to quantitatively depict the intra-graft flow in both arterial and venous grafts. We initially demonstrated the flow values of saphenous vein grafts to the three major coronary arteries.

METHODS

       Patient population. From October 1998 to January 2001, one hundred and twenty-three patients (93 men and 30 women; mean age, 63.2 ± 10.9 years; age range, 34–85 years) who had undergone CABG surgery were scanned using EBT for the evaluation of CABG status. These consecutive patients were selected and eligible for the study. For evaluating the patency or occlusion of CABG after bypass surgery, physicians referred EBT studies in all of our patients.

Figure 1

Schematic diagram illustrates use of density (in Hounsfield units; HU) versus time (seconds) curves in deriving blood flow parameters on electron-beam tomography flow study. The flow characteristics of the ascending aorta and coronary artery bypass graft (CABG) can be evaluated and compared by using the parameters of peak CT, peak time, rise CT/rise time, pass time and blood flow values. Blood flow is proportional to (ACABG/AAO)/(TCABG-TAO) (ACABG, cross-hatched region area under the CABG flow curve; AAO, stippled region area under the aortic curve; TCABG and TAO are centers of gravity for CABG curve and for aortic curve).

The duration between bypass surgery and EBT study was 10.5 ± 18.2 months (range, 0.03–78.5 months). Patients who had recurrent symptoms of myocardial ischemia or who were suspected to have CABG occlusion received post-operative conventional angiography. Post-operative cardiac catheterizations were done within 2–10 weeks (39.5 ± 21.2 days) before or after EBT scanning in 26 patients. According to the operative reports, there was a total of 365 CABG, including 54 left internal mammary arteries (LIMA), 2 right internal mammary arteries (RIMA) and 309 saphenous vein grafts (SVG). There were 131 grafts (35.9%) bypassed to the left anterior descending (LAD) coronary artery (including diagonal and ramus intermedius branches); 126 grafts (34.5%) to the left circumflex (LCX) coronary artery (including obtuse marginal branches), and 108 grafts (29.6%) to the right coronary artery (RCA) system.

       EBT flow protocol. The multi-slice mode (MSM) of flow study was performed on EBT C-150XP scanner (Imatron, South San Francisco, California). All patients underwent EBT examination headfirst in a supine position perpendicular to the gantry. Electrocardiographic (ECG) leads were attached to the skin in the right and left infraclavicular areas and left lower thorax. ECG triggering was used at 0% R-R interval (R wave) of the cardiac cycle.
       A preview scanning was initially performed to ensure that the uppermost image would be taken at the aortic arch level. In the patients with an LIMA or RIMA, the uppermost image was at the thoracic inlet. Multi-slice flow protocol was selected from the EBT scanning menu and 4 tungsten targets were used. Eight transaxial sections with 10 frames of each section were acquired. Section thickness was 8 mm, with a 4 mm gap from each section. The image matrix was 256 x 256 and field of view was 18 cm, yielding a pixel size of 0.7 x 0.7 mm2. The acquisition time was 50 milliseconds (msec) per slice with an 8 msec interval from each slice. The total scan time depended on the patient’s heart rate (the triggered acquisition sequence was initiated on the R wave of every third heartbeat). The scanning began with the contrast agent injection. A 16 or 18 gauge plastic catheter was placed into a right or left antecubital vein. Non-ionic contrast medium (Ultravist 300, Schering, Berlin, Germany) was then injected with a power injector (Medrad 100, Pittsburgh, Pennsylvania) at a rate of 8 ml/second for a total dose of 35 ml.

Figure 2a

       EBT data processing and analysis. Two experienced readers who were unaware of both operative reports and post-operative cardiac catheterization findings separately evaluated the flow datasets on EBT consoles. All flow parameters were derived from analyzing their time-density curves (TDC) or gamma-variate curves (Figure 1). The same levels of flow images were employed by the two observers to evaluate study agreement. The criteria for graft patency, based upon our experience and other studies,1–8 was that the graft could be visualized on at least two non-contiguous levels or that a good TDC could be created. Occlusion of a graft was diagnosed if it could not be visualized or its TDC was flat in comparison with that of the ascending aorta.

       Comparison of EBT and post-operative angiography. Twenty-six patients underwent post-operative CABG angiography using the conventional Judkins techniques. If a patent graft or closed stump could not be visualized, an aortic root power injection or ultra-selective angiography of the CABG in one or more projections was carried out. Two investigators analyzed the angiographic films and made the diagnosis. The results of EBT flow study were compared double-blindly with the conventional angiography. Sensitivity, specificity and accuracy values were obtained using the traditional equations. Interobserver variability was performed.

Figure 2b

(A) Conventional post-operative angiography identified a widely patent saphenous vein graft (SVG, white arrow) to the left anterior descending coronary artery (LAD, black large arrow). The back-flow and distal anastomosis to the LAD was visualized well (small black arrow). (B) Electron-beam tomography flow study showed that the time-density curve of this patent SVG (marked with square) was morphologically similar to the curve of the ascending aorta (marked with A).

       Statistical analysis. The ability of the flow study to correlate with angiographically confirmed bypass graft patency or occlusion was evaluated by calculating sensitivity, specificity and accuracy; positive and negative predictive values were also investigated using 2 x 2 contingency tables. Comparisons of flow variables among the ascending aorta, arterial and venous grafts, and comparisons of flow differences of SVG to LAD, LCX and RCA were carried out using the one-way analysis of variance test (ANOVA). The unpaired Student’s t-test was also performed, where p < 0.05 was defined as significant. Cohen’s Kappa statistic was used to compare the inter-observer agreement. Whenever possible, data were presented as the means ± one standard variation.

RESULTS

       Procedural success of EBT flow study. The EBT procedures were tolerated well without complications in all patients. Twelve flow studies (9.8%) were excluded due to impaired image quality, such as poor contrast injection (no contrast in coronary vessels), unpredictable flow images caused by poor image acquisition during the scanning, rather than due to the occlusion of the graft. The diagnosis of other grafts could be made on the basis of good contrast enhancement. However, if the contrast enhancement was adequate and the CABG was not visualized, then this CABG was diagnosed with occlusion. From analyzing flow curves, twenty-one out of 56 time-density curves (37.5%) of internal mammary artery (IMA) grafts could not be obtained because of small vessel size and image artifacts, such as metallic vessel clips, cardiac motion, bad electrocardiographic triggering and partial volume effect. The visualizations of SVG were more effective than IMA grafts, and more time-density curves of SVG (265 of 309 grafts; 85.8%) were available to be interpreted than IMA grafts (p < 0.05). The uninterpretable TDCs of SVG included 3 out of 75 (4.0%) to the LAD system (including diagonal and ramus intermedius branches), twenty-six out of 126 (20.6%) to the LCX system (including obtuse marginal branches) and 15 out of 108 (13.9%) to the RCA system (including the posterior descending artery). The reason for the lower rate of uninterpretable SVG to the LAD system was not clear.

       Accuracy of EBT flow study. By comparison with post-operative cardiac catheterization in 26 patients, fourteen of 16 angiographically closed grafts and 68 of 75 patent grafts were correctly identified by EBT flow study, yielding a sensitivity of 87.5% and specificity of 90.7%. Overall accuracy for EBT flow study was 90.1%. The 2 false-negative interpretations (occluded graft was misinterpreted as patent) by flow study, and the 7 false-positive interpretations (patent graft was misinterpreted as occluded), yielded a positive predictive value and negative predictive value of 66.7% and 97.1%, respectively. The two false-negative interpretations occurred in SVG grafts, caused by some high-density calcified lesions. The false-positive interpretations were mostly caused by suboptimal image resolution. The Cohen’s Kappa statistic for inter-observer agreement of flow study was 0.62. Two angiographically confirmed patent but minimally narrowed (< 50% stenosis) venous grafts and two significantly narrowed (> 50% stenosis) venous grafts were not correctly interpreted on EBT flow studies, because these 4 stenosed venous grafts showed similar flow patterns to non-stenosed grafts.

       The characteristics of intra-graft flow.A wide patent bypass graft has an excellent time-density curve with its rise and fall morphologically similar to that of the ascending aorta, and an occlusion has no flow curve or the curve is flat (no significant rise and fall) (Figure 2). The intra-graft flow parameters of the IMA, SVG and ascending aorta are presented in Table 1. The values of flow parameters of patent grafts (either IMA or SVG) were significantly lower than those of the ascending aorta (p < 0.001), especially in peak CT, rise CT/rise CT time and blood flow, whereas the pass time was significantly longer than in the ascending aorta (p < 0.01). There were no significant differences in peak CT times between CABGs and the ascending aorta (p > 0.05). The contrast attenuations and flow values of the SVG were significantly higher than of the IMA (p < 0.01 and p < 0.001, respectively); however, the pass time of the SVG was shown to be shorter than that of the IMA (p > 0.05). These results confirm that the flow velocity inside the SVG was faster than the IMA, but not statistically different (p = 0.17). The peak CT values of patent IMAs and SVGs were on average 53.0 ± 17.7% and 59.8 ± 15.5% of the ascending aorta, and the intra-graft flow values were 31.6 ± 20.4% and 39.4 ± 21.9% of the ascending aorta, respectively.
       The SVG flow data based upon their distributions to the three major coronary arteries are presented in Table 2. The blood flow characteristics, such as peak CT value, peak time, rise CT/rise CT time and pass time of SVG to LAD, LCX and RCA systems, were similar to each other (p-values > 0.05). The peak times and pass times of RCA grafts were the smallest, followed in an increasing order by the LCX and LAD, while the intra-graft flow values of the RCA were larger than those of the LCX (p > 0.05), and significantly larger than those of the LAD (p < 0.05). These findings demonstrate that the intra-graft flow velocities are greatest in the RCA, followed by the LCX and LAD. A marked trend was found toward higher intra-graft flows of non-stenosed venous graft (6.9 ± 2.8 ml/min/g) than in stenosed venous grafts (4.2 ± 1.3 ml/min/g); however, this result was not significant because fewer angiographically confirmed stenotic lesions were evaluated in this group.
       Two experienced observers measured and compared the flow values (ml/min/g) of each bypass graft to identify the flow study reproducibility. The interobserver variability [(Observer 1 – Observer 2)/mean of Observer 1 and Observer 2] is presented in Table 3. The lowest variability occurred in the ascending aorta, followed in increasing order by the IMA, SVG to LAD, SVG to LCX and SVG to RCA.

DISCUSSION

       The utilization of EBT multi-slice flow study to detect the patency of CABG has been reported as early as 1986.9 Subsequent investigations showed that EBT flow protocol was a very valuable method for detecting bypass graft occlusion.4,5,10 Although the criteria for the assessment of graft patency were well established in these former studies, no flow data were obtained to quantitatively evaluate the graft patency. In other experimental studies, EBT flow protocols were successfully used to quantify regional myocardial perfusion11,12 and to measure regional cerebral blood flow in dogs.13 The theoretical aspects of flow equations were also discussed in these primary experimental or clinical studies.14,15
       Basically, using the time-density curve of EBT flow protocol, absolute flow can be calculated for any blood vessels as the ratio of peak enhancement of the time-density curve in that region to the area under the aortic or left ventricular time-density curve, that is F/V = P/A [F/V equals blood flow per unit volume or gram of vessel (ml/min/ml or ml/min/g; blood has a density of 1.05 g/ml), P equals peak of the vessel flow curve, and A equals the area under the aorta or left ventricle time-density curve]. Our study was designed on the basis of this theory. Thus, all the flow parameters of CABG could be obtained and compared with aortic time-density curves on EBT flow study.

       Criteria for graft patency. Our criteria for evaluating CABG patency were based on visualization of the graft on at least two non-contiguous cross-sectional levels or that the graft had an excellent time-density curve morphologically similar to that of the ascending aorta. Our criteria for diagnosing CABG occlusion were that the graft could not be visualized and the graft had no significant flow curve or the curve was flat without rise and fall.
Using these criteria, we demonstrated that the sensitivity, specificity and accuracy of EBT flow study for the assessment of CABG occlusion were 87.5%, 90.7% and 90.1%, respectively. These values were very similar to the study by Stanford (sensitivity, specificity and accuracy were 93.4%, 88.9% and 92.1%, respectively),4 but lower than the study by Bateman (96%, 97% and 96%, respectively).5 In this study, we initially obtained the intra-graft flow values of both IMA and SVG grafts, and we also made further considerations of the difference of SVG blood flow with its distribution to the three major coronary arteries. We did not use the flow values as our diagnostic criteria for assessing CABG patency or narrowing. More angiographically confirmed patients were needed to set the quantitative criteria for the evaluation of CABG patency and graft stenosis.

       Arterial versus venous intra-graft flow. By analyzing quantitative flow data on time-density curves, we demonstrated that the peak CT value and flow value of a patent IMA graft were 53.0 ± 17.7% and 31.6 ± 20.4% of the ascending aorta, respectively. For a patent SVG, the percentages were 59.8 ± 15.5% and 39.4 ± 21.9%, respectively. The lower peak CT values of CABG as compared to the ascending aorta were probably caused by the smaller vessel size of CABG and more partial volume effect on CABG (8-mm slice thickness scanning). Our data showed that the average flow value of the SVG was higher than of the IMA graft, while the average peak time and pass time of the SVG were lower than the IMA. These results confirmed that the flow velocity of the SVG was faster than the IMA. Rumberger used DPT (the time difference between peak opacification in the aorta and the bypass graft) to assess the intra-graft flow velocity, and found a high correlation with the Doppler velocity study.16 In our study, we directly used the parameters of flow value, peak time and pass time to quantify bypass graft flow velocity and flow value.

       Protocol and study limitations. The EBT flow study has suboptimal spatial resolution (maximal in-plane resolution is 0.59 ± 0.59 mm2, slice thickness is 8 mm, with a 4 mm gap between every two levels); thus, the entire course of the bypass graft and anastomoses could not be visualized. The measurements of intra-graft blood flow could be affected by the slice thickness during scanning, course of the graft through the slice level and partial volume effects. A major limitation of the flow study is that partially obstructed (< 50% stenosis) versus non-obstructed grafts cannot be interpreted, and the status of sequential distal anastomoses cannot be determined. Moreover, flow values of some grafts, especially in the IMA, cannot be obtained due to unfit time-density curves. The enhanced capabilities of the EBT scanner permit acquisition of the high-resolution EBT images, which can be reconstructed three-dimensionally to show the entire course of the graft.17 High-resolution three-dimensional images were not obtained in this study.
       The interobserver agreement of EBT flow study was relatively low, with k = 0.62 in our group. A greater percentage of patients with angiographic correlation would have been desirable.
Conclusion

       In spite of the above limitations, multi-slice flow study of electron-beam tomography is currently used clinically and is a relatively accurate noninvasive method for the assessment of coronary artery bypass graft patency4,5 and intracoronary artery stenting.18 It not only identifies patency versus occlusion of either an internal mammary artery graft or a saphenous vein graft, but also quantifies the intra-graft flow of a patent bypass graft.

1. Ha JW, Cho SY, Shim WH, et al. Noninvasive evaluation of coronary artery bypass graft patency using three-dimensional angiography obtained with contrast enhanced electron beam CT. Am J Radiol 1999;172:1055–1059.
2. Dai RP, Zhang SX, Lu B, et al. Electron-beam CT angiography with three-dimensional reconstruction in the evaluation of coronary artery bypass grafts. Acad Radiol 1998;5:863–867.
3. Achenbach S, Moshage W, Ropers D, et al. Noninvasive, three-dimensional visualization of coronary artery bypass grafts by electron beam tomography. Am J Cardiol 1997;79:856–861.
4. Stanford W, Brundage BH, Macmillan R, et al. Sensitivity and specificity of assessing coronary bypass graft patency with ultrafast computed tomography: Results of a multicenter study. J Am Coll Cardiol 1988;12:1–7.
5. Bateman TM, Gray RJ, Whiting JS, et al. Prospective evaluation of ultrafast cardiac computed tomography for determination of coronary bypass graft patency. Circulation 1987;75:1018–1024.
6. Stanford W, Krachmer M, Galvin JR, et al. Ultrafast computed tomography in assessing coronary bypass grafts. Am J Card Imaging 1991;5:21–28.
7. Hoogendoorn LI, Pattynama PMT, Buis B, et al. Noninvasive evaluation of aortocoronary bypass grafts with magnetic resonance flow mapping. Am J Cardiol 1995;75:845–848.
8. Wintersperger BJ, Engelmann MG, Smekal AV, et al. Patency of coronary bypass grafts: Assessment with breath-hold contrast-enhanced MR angiography — Value of a nonelectrocardiographically triggered technique. Radiology 1998;208:345–351.
9. Bateman TM, Gray RJ, Whiting JS, et al. Cine computed tomographic evaluation of aortocoronary bypass graft patency. J Am Coll Cardiol 1986;8:693–698.
10. Stanford W, Galvin JR, Thompson BH, et al. Non-angiographic assessment of coronary artery bypass graft patency. Int J Card Imaging 1993;9:77–86.
11. Rumberger JA, Feiring AJ, Lipton MJ, et al. Use of ultrafast computed tomography to quantitate regional myocardial perfusion: A preliminary report. J Am Coll Cardiol 1987;9:59–69.
12. Wolfkiel CJ, Ferguson JL, Chomka EV, et al. Measurement of myocardial blood flow by ultrafast computed tomography. Circulation 1987;76:1262–1273.
13. Gobbel GT, Cann CE, Iwamoto HS, et al. Measurement of regional cerebral blood flow in the dog using ultrafast computed tomography: Experimental validation. Stroke 1991;22:772–779.
14. Gobbel GT, Cann CE, Fike JR. Measurement of regional cerebral blood flow in the dog using ultrafast computed tomography: Theoretical aspects. Stroke 1991;22:768–771.
15. Rumberger JA, Bell MR. Measurement of myocardial perfusion and cardiac output using intravenous injection methods by ultrafast (cine) computed tomography. Invest Radiol 1992;27:S40–S46.
16. Rumberger JA, Feiring AJ, Hiratzka LF, et al. Quantification of coronary artery bypass flow reserve in dogs using cine-computed tomography. Circulation Res 1987;61(Suppl II):II117–II123.
17. Lu B, Dai RP, Jing BL, et al. Evaluation of coronary artery bypass graft patency using three-dimensional reconstruction and flow study on electron beam tomography. J Comput Assist Tomogr7 2000;24:663–670.
18. Lu B, Dai RP, Jing BL, et al. Detection and analysis of intracoronary artery stent after PTCA using contrast-enhanced three-dimensional electron beam tomography. J Invas Cardiol 2000;12:1–6.