- Email: [email protected]
Assessment of coronary angioplasty by an automated digital angiographic method
Assessment of coronary angioplasty by an automated digital angiographic method Digital subtraction coronary angiograms (DSA) of 63 patients who had undergone coronary angioplaaty (PTCA) for a total of 73 lesions were analyzed with an aUtOmated border-detecting computer program capable of simultaneous geometrfc and dendtometrtc crose-eectional 8rea estimation. The computer measurements were compared with visual interpretation of the 35 mm cinesngiograms. The results Indicated that visual reports of cineangtograms tend to overesttmete the pre-PTCA diameter percent stenosis and to underestimate the post-PTCA res+dusl &tenosis in comparison with the computer (p < 0.001 In both cases). There WBS good agreement between geometric and densitometric area percent stenoses calculated by the program on the pre-PTCA digital angiograms (f = 0.82, p < 0.001, mean of their differences = -0.2 with standard deviation = 8.1). Following PTCA, however, important discrepancies between the two methods existed (r = 0.71, p < 0.001, mean of their differences = 1.0 with standard deviation = 18.8). Following PTCA (but not pre-PTCA), densitometric evaluation demonstrated 8 significantly greater mean coefficient of variation between different views (69%) than did the geometric evaluation on the same views (24%). We conclude (1) that visual interpretation of tine coronary angiograms compares poorly with quantitative methods for both the selection of PTCA candidates and the assessment of the results; (2) that the geometric and densitometric characteristics do not agree in describing the degree of post-PTCA residual stenosis; and (3) that after angioplasty, important discrepencies between densitometric evaluation in different views are observed. (AM HEART J 1988;116:1181.)
Demosthenis Katritsis, Ian C. Cooper, MRCP,
MD, David A. Lythall, MB, BS, Mark H. Anderson, MRCP, and Michael M. Webb-Peploe, FRCP. London, England
Subjective visual assessment of percent stenosis of coronary arteries from routine cineangiograms is subject to large intra- and interobserver variability,’ and a poor correlation between angiographic and postmortem pathologic findings has been demonstrated.* Previous studies3 have also shown that the minimal cross-sectional area of a stenosis rather than its percent diameter reduction is the better determinant of its hemodynamic significance. In order to assess more precisely the functional significance of coronary artery stenoses, computer-aided measurements of the geometric dimensions of the stenosis4*5 and densitometric evaluation of its crosssectional area have been used.4,6* 7 Analysis of coronary angioplasty (PTCA) results presents particular problems. It has been showns,s that in severe lesions visual estimation tends to overstate the severity of the pre-PTCA obstruction. From
the Department
Supported Received Reprint Hospital,
of Cardiology,
by the Cardiac for publication requests: London
Research Feb.
5, 1988;
D. Katritais, MD, SE1 7EH, England.
St. Thomas’s Fund,
accepted Dept.
Hospital.
St. Thomas’s July
Hospital. 1, 1988.
of Cardiology,
St. Thomas’s
PTCA usually causes intimal tearing and gross di$ortion of the angiographic appearance of the artery.lO* l* Important discrepancies between geometric and densitometric measurements of post-angioplasty lumens may therefore occur, and densitometric analysis in a single view has been proposed for the evaluation of PTCA results.‘* In other studies, however, densitometry has not been found significantly different from edge detection,13 and an impressive lack of correlation between densitometric measurements in orthogonal views, both pre- and post-PTCA, has been demonstrated.14 All these reports,i2-14 however, have been based on relatively small numbers of cases. In this study we applied an automated computer program capable of simultaneous geometric and densitometric analysis of digital angiograms (DSA) for evaluation of the results of PTCA. The aims were to: (1) compare these measurements with the visual interpretation of routinly reported 35 mm cineangiograms; (2) test the validity of densitometric analysis in a single angiographic projection for the assessment of pre- and post-angioplasty stenoses; and (3) compare the geometric and densito1181
1182
Kntritsis
et nl.
American
Table
I. Coronary
artery
lesions dilated
by PTCA
No. of patients Total lesions LAD RCA
LAD, left anterior SVBG. saphenous
Fig. 1. Digital subtraction angiograms showing geometric and densitometric calculations before (top panel) and after angioplasty (bottom panel).
metric characteristics cedure.
of the lumens after the pro-
METHODS Imaging system. Angiograms were obtained on a realtime digital image acquisition and processing system (Digitron 2, Siemens AG, Munich, W. Germany) interfaced to a conventional cineangiographic unit (Pandoros Generator/Cardioscop U, Siemens AG, Erlangen, W. Germany). An image module with 20 megabyte semiconductor memory was used for acquisition, and postnrn~nr.,.“:r,- ...ua ,,,*X uui on a videu processor equipped with a 168 megabyte Winchester disk drive. Images were acquired at 8 frames/set in a 512 X 512 matrix with lo-bit depth. Prior to digitization, logarithmic preamplification
November 1988 Heart Journal
63 73
cx
34 15 13
SVBG
11
descending; vein bypass
RCA. graft.
right
coronary
artery;
Cx, circumflex;
compensated for nonlinear x-ray absorption resulting from different object density and thickness (LambertBeer law). Phasic mask subtraction was used, i.e., diastolic mask frames were used for subtraction from diastolic images and systolic masks were used for systolic images. Cineangiograms at the same views were acquired at 25 frames/set. Cineangiograms were routinely reported by the cardiologist who had performed the angioplasty and who was unaware of the DSA measurements. Analysis of the DSA images was performed on those views (one to four) in which the pre-PTCA lesion was clearly seen and the same views were used for the postPTCA analysis. In this way 58 lesions were studied in two to four projections and 15 lesions were studied in one projection only. Individual images were selected from both the pre- and the post-PTCA recordings and were subsequently stored on the Winchester disk. An image was selected for analysis that met the following criteria: (1) The image was selected from such a phase of the contrast injection that both the lesion and the adjacent normal segments of the artery were equally opacified. (2) All segments had to be displayed without foreshortening or overshadowing by side branches or other anatomic structures. (3) The compared images were selected at identical times of the cardiac cycle (late diastole). The same views and similar images were selected for the pre- and post-angioplasty recordings. For the quantitative studies, each image was recalled on the monitor and was submitted to geometric and densitometric analysis performed by a computer program based on a borderdetecting dynamic search algorithm.15 Edge detection program. This program is based on a modular, fully automated border detection algorithm developed by Pope, Parker, et al. and described in detail elsewhere. I5 Having displayed the preselected image on the monitor, the operator, with a light-pen cursor, defines the search area the program will use to find the vessel edges and the complete vessel length to be analyzed. Geometric cross-sectional area at each point along the centerline is obtained by assuming circular cross-sections and densitometric cross-sectional area is obtained by integrating the density across the perpendicular profile from edge to edge. The density values are background corrected by subtracting an average background value determined by the densities beyond the edges of the vessel. Two pairs of cursors representing the largest and smallest values are then automatically placed over the
Volume Number
Table
116 5, Part 1
II. Comparison
PTCA
of computer
estimations
on subtracted
and unsubtracted
assessment
Densitometric
area % stenosis
Values are expressed as mean + SEM. r, correlation coefficient; Sub, subtracted;
1183
Post-PTCA Unsub
Sub
Geometric area % stenosis
DSA
images
Pre-PTCA
Diameter % stenosis
by computed
Sub
58.3 + 3.5 59.7 * 3.3 (r = 0.90, p < 0.001) 81.4 + 2.8 82.5 f 2.9 (r = 0.88, p < 0.001) 80.0 f 3.1 80.1 + 3.6 (r = 0.89, p < 0.001)
Unsub
36.3 k 6.5 36.7 (r = 0.97, p < 0.001) 55.1 -+ 9.2 55.4 0. = 0.97, p < 0.001) 42.6 I+ 9.7 43.9 (r = 0.96, p < 0.001)
k 6.8 5 9.5 + 9.4
Unsub, unsubtracted.
vessel and their position is further adjusted by the operator. Percent diameter stenosis, percent geometric crosssectional area stenosis, and percent densitometric crosssectional area stenosis are then automatically displayed (Fig. 1). The stenotic and reference areas are reproduced on the post-PTCA images. For statistical analysis, the measurements obtained in all the angiographic projections studied were averaged for each method and every lesion. Calibration of the digital images can be achieved by using the inside diameter of the catheter guiding system as reference. Strictly, absolute calibration of the system is unnecessary when comparing relative percentage stenoses before and after PTCA. Such comparison of relative stenoses also avoids potential errors from pincushion distortion, causing selective magnification of the peripherally located guiding system. Subtracted and unsubtracted images. Automated border detection was applied to 13 projections of 6 lesions in both subtracted and image (unsubtracted) modes and the results were compared. In two instances, border detection on unsubtracted images was not possible; the lesion was overshadowed by either the pulmonary artery (pacing) catheter or, in the case of a caudal view, by the spine. In the rest of the projections, geometric and densitometric stenoses derived from the two image modalities were similar with a very good correlation (presented in the Results section). Consequently, only subtracted images were selected for further analysis. Statistical analysis. Student’s two-tailed paired t test and linear regression analysis were used for the comparison of stenoses as assessed by different methods. In addition, the mean and the standard deviation of their differences were considered. Because some data were not normally distributed (i.e., visual reports), nonparametric statistical tests were additionally employed (Wilcoxon signed-rank test and Spearman’s rank correlation). The results, however, did not differ significantly from those of parametric analysis. Thus parametric results only are presented in the text. The coefficient of variation, as a percentage, was used to assess the value of taking multiple views for densitometric analysis. Patients. Sixty-three patients who had undergone successful PTCA were studied. PTCA was performed on 73 lesions that were thought to be responsible for exercise-
induced myocardial ischemia demonstrated by thallium201 scintigraphy. The following categories were not included in the study. (1) Cases with obvious postangioplasty dissection interfering with the program’s edge detection. (2) Patients who had previously undergone coronary artery bypass grafts and had angioplasty of their native vessels. The native coronaries were usually distorted due to the grafts, so it was difficult to define the reference (normal) part of the vessel. (3) Lesions that could not. be displayed without their being overshadowed by side branches, whatever projection was used. The sites of the studied lesions are summarized in Table I. All lesions were evaluated by means of visual assessment and combined geometric and densitometric quantitation by the automated computer program. RESULTS Subtracted
statistically
vs unsubtracted images. There was no significant difference in diameter, geo-
metric area, or densitometric area between subtracted and unsubtracted
percent
stenoses
images, either pre- or post-PTCA. A very good correlation existed between the two image modalities, whatever kind of measurement was considered (Table II). Pre-PTCA VI post-PTCA results. The percent stenoses as evaluated by the various methods both before and after PTCA are shown in Table III. There was a significant improvement in luminal diameter after PTCA (p < O.OOl), whatever method was used. Visual
reports
vs computer
diameter
measurements.
The visual reports of cineangiograms significantly overestimated the pre-PTCA diameter percent stenosis and underestimated the post-PTCA residual stenosis in comparison with the computer program (p < 0.001 in both cases). Fig. 2 and Table IV show the important disagreement between the two methods and the virtual lack of any correlation between them.
Before PTCA, a total of 43 out of the 73 lesions dilated (i.e., 59%) were estimated to represent less than 60% diameter stenosis by the computer. Sev-
November
I 184
Katritsis
et al.
American
Table III. Comparison of pre- and post-PTCA estimated visually and by the computer Pre-PTCA
Values
are expressed
as mean
Pre-PTCA
12.7 ” 1.7
55.3 i 1.1
23.2 i 1.8
78.8
+ 1.1
37.9 + 2.7
79.0
+ 1.2
36.9
r P
r, correlation
P
Mean of differences SD of differences r, correlation
coefficient;
Post-PTCA 0.48
0.04 0.76 26.3
-Co.001
-10.5 15.0
13.8 SD, standard
deviation.
enteen lesions (i.e., 23%) were estimated to be less than 50%. All 76 lesions were reported to be more than 60% stenotic before PTCA on visual estimation. Geometric vs densitometric computer measurements. There was no statistically significant difference between geometric and densitometric area stenoses, either pre-PTCA or post-PTCA. The two image modalities yielded variables that were well correlated pre-PTCA (r = 0.82, p < O.OOl), whereas there was a modest correlation post-PTCA (r = 0.71, p < 0.001) (Fig. 3 and Table V). However, analysis of their differences indicated important discrepancies post-PTCA (standard deviation of the differences = 18.6) that did not exist pre-PTCA (standard deviation of the differences = 6.1). Multiple vs single projections; metric evaluation. For those
densitometric
and geo-
58 lesions studied in more than one projection, the coefficient of variation was calculated for the degree of densitometric area stenosis assessed from multiple views of each individual lesion. The mean of the coefficients of variation for all lesions was 8% pre-PTCA and 69 % post-PTCA. This indicates a wide variation of den-
oitr\mntr:n
-7,17-n“c-uIY
~$t~i&
f;cm
the
same
coefficient;
SD, standard
Post-PTCA 0.71
18.6
deviation.
k 3.0
Table IV. Correlation between visual reports and computer measurements of percent stenoses of luminal diameter
r
0.82
Mean of differences SD of differences
i SEM.
Pre-PTCA
1988 Journal
Table V. Correlation between geometric and densitometric evaluations of percent of area stenosed
Post-PTCA
81.6 i- 1.2
PC, Diameter stenosis (visual report) rC1 Diameter stenosis (computer program) “;# Geometric area stenosis (computer program) 5, Densitometric area stenosis (computer program)
stenoses
Heart
lesion
in
different gngiographic projections after PTCA, but not before. In the case of geometric area percent
stenosis, the mean of the coefficient of variation for all lesions was 7% pre-PTCA and 25% post-PTCA. Therefore post-PTCA, geometric areas do not show the same degree of variation demonstrated by densitometry. DISCUSSION
The value of quantitative methods in assessing the physiologic significance of coronary artery disease and its response to therapeutic interventions has been well documented.**9s’3*14 Geometric methods assume either a circular (for one-plane views) or elliptical (for biplane views) cross-sectional shape of the lesion. To avoid geometric assumptions that do not characterize asymmetric lesions, densitometric techniques have made use of the concept that in an ideal setting the density across an artery in a radiographic image is proportional to the crosssectional area at that point.‘j Several studies6*7 have proposed that even a single radiographic projection could be sufficient for the estimation of the degree of stenosis. However, densitometric evaluation is based on several premises.6 First, it is assumed that radiographic magnification and exposure are uniform. Second, the distribution of the dye must be equal in both normal and stenotic arterial segments. In clinical practice this is not always the case. Digital subtraction angiography. Since almost all quantitative methods require digitization of conventional cineangiograms, on-line digital acquisition of coronary arteriograms is ideal for such applications. Cumbersome digitization of cineangiograms is avoided and the need for high frame rate angiographic views is reduced. Suppression of distracting background anatomy and enhancement of the contrast level can significantly improve the yield of diagnostic information. Disadvantages of subtraction angiography include inferior spatial and temporal resolution compared with cineangiograms.16 Furthermore, misregistration artifacts caused by respiratory movement remain one of the most common causes of image degradation when conventional
Volume
116
Number
5, Part 1
PTCA assessment Pre-PTCA
by computed DSA
Post-PTCA
r = 0.04
r = 0.48
T.5 90.P 0” c 80-
n
d . f 70fi; E .m 60 cl El 2 soE 9
1185
q q n
II q q q
. 40m 30:.
, 30
.
,
40
.
50
,
.
60
Visual Report
,
q ,
,
70
.
,
80
Diameter
,
90
Stenosis
(
-20 1,
100
-20
(%)
0
20
Visual Report
40
60
Diameter
80
Stenosis
100
(%)
Fig. 2. Relationship between visual estimation and computer measurements of percent stenosesof luminal diameter.
Pre-PTCA 100
y = 4.76 + 0.94x
r = 0.82
70-
$
40-
$
zo-
.s
Oi? . .s2 -20 E . 8 -4o-
60-
$ (ij
r = 0.71
‘;i 80a. i 60e .
-2 $
y = 7.35 + 0.78x
100 -I
1
7 5 go.I? 0” 5 so(I) zi 2
Post-PTCA
50-
q
q 40
. 40
I
.
50
Geometric
I 60
I
, 70
.
I 80
Area Stenosis
.
, 90
.
, 100
(“I&)
-60
. -60
, -40
.
, -20
Geometric
.
, 0
.
, 20
.
, 40
Area Stenosis
.
, 60
.
, 80
.
( 100
(%)
Fig. 3. Linear regression analysis of geometric and desitometic evaluations of percent of area stenosed.
first-order subtraction is employed. Newer methods such as recursive filtering and “second-order” techniques such as hybrid subtraction should eliminate these problems.16 Recently, Mancini et al.” reported the successful evaluation of a similar DSA program by studying well-defined lesions created by means of cylinders of known diameter placed over canine coronary arteries. This technique allowed accurate luminal diameter measurements, and direct comparison of digital and tine images showed some deterioration in both accuracy and reproducibility when films were used.
Tn vit,ro evaluation of a similar method on human cadaver coronary arteries” showed very good correlation of the geometric and densitometric results with histopathologic measurements. These results indicate that on-line digital images can provide greater morphologic precision than routinely processed cineangiograms. Assessment of PTCA results. Analysis of the results of PTCA presents particular problems. Meier et a1.,g found that the intraobserver coefficient of variation was significantly reduced when the mean estimations from three projections instead of the projec-
November
1166
Katritsis
et al.
tion showing the most severe stenosis were used. Bove et al.* demonstrated that visual assessment overestimated the percent diameter stenosis of severe lesions. Serruys et a1.,12in a study of 18 cases by quantitative methods on conventional angiograms, found a good agreement between the densitometric precent area stenosis and circular percent area stenosis before PTCA. After PTCA, important discrepancies were observed between the two types of evaluation. These investigators suggested that the mechanical disruption of the internal artery wall caused by the procedure leads to eccentric, asymmetric morphologic changes that cannot be accurately assessed from the detected contour of the vessel when a single angiographic view is used. They therefore proposed the use of densitometry as essential for that purpose. However, in a recent digital evaluation of angiographic study, l3 densitometric both the pre- and post-PTCA lumens was not significantly different from the edge detection method, and the correlation between measurements in orthogonal projections was similar for both techniques either before or after PTCA. Sanz et a1.,14in a study of 13 cases with a digital method similar to our own, showed a lack of correlation between densitometric measurements in orthogonal views both before and after PTCA. This study represents the first extensive clinical application of an automated computer-aided technique for digital image quantitation before and after PTCA. It allowed rapid quantitative assessment of coronary artery stenoses both before and after the intervention. The results support the evidence for inaccuracy of visual assessment of coronary stenosis both pre- and post-PTCA. In accordance with previous reports,s we have found that this overestimation is worse with more severe stenoses (i.e., pre-PTCA). For example, the computer found 23 % of all lesions to have a stenosis of less than 50% of the lumen diameter. Previous reports’g’20 have shown that reduction of the arterial lumen causes significant alteration in resting flow only when the stenosis is greater than 85 % , and that lumen stenosis may limit significantly maximal flow response to hyperemic stimuli when it is greater than 45%. Recently, Doppler catheter studieszl for the assessment of coronary flow reserve demonstrated that maximal hyperemic coronary flow is usually impaired by stenoses affecting more than 70% of the luminal area or by those producing more than 50% diameter obstruction. Furthermore, it has been shownz2 that U~IgLpiasLy or stenoses of less than 60% carries a substantial risk of complications and is better avoided. The relevance of these findings to the
American
1988
Heart Journal
selection of PTCA candidates is obvious. Overestimation of diameters of less than 1 mm is a welldescribed feature of automated quantification techniques.16 Phenomena such as motion blurring, limited spatial resolution, and geometric magnification all promote overestimation of diameters at this level. Mancini et a1.,17however, accurately measured diameters as low as 0.8 mm, and pre-PTCA diameters averaging 1.0 mm8 or 1.28 mm12 are the general rule. Such overestimation would presumably affect very tight lesions and not the 50% to 60% stenoses where our visual report and the computer disagreed. The program’s geometric (cross-sectional area) assessment of the pre-PTCA stenoses correlated well with its densitometry derived values and the discrepancies between the two methods do not seem to be of clinical importance. In the post-PTCA assessment, however, discrepancies were noted between the two methods, in accordance with the findings of previous reports.12 In contrast to the report of Sanz et a1.,14 before PTCA both densitometric and geometric analysis demonstrated low mean coefficients of variation in the projections best demonstrating the lesion. PostPTCA, however, the densitometric variation in the same views was much larger than the geometric variation, suggesting that one projection is not enough for densitometric analysis. It has been assumed that following angioplasty, distortion of the vessel lumen in most cases makes any accurate morphologic description difficult. The fact that this degree of variation does not exist in geometric evaluation post-PTCA suggests that the densitometric discrepancies between views cannot be ascribed to the irregular anatomy of the postangioplasty lumen alone. Similarly, inaccuracies of the densitometric method itself do not appear to provide an adequate explanation, since this variation did not exist pre-PTCA. It may be that the distortion of the vessel interferes with the mixing of the radiographic contrast medium and blood, hence invalidating any assumptions about dye distribution. Quantitative methods that use edge dection programs seem to be accurate and reproducible,14 and may have both diagnostic and prognostic significance. Stenotic diameters or cross-sectional areas acquired in this way have been found to indicate with reasonable accuracy the consequent reduction in coronary flow capacity21,23 and to predict reocclusion rates or left ventricular function recovery following thrombolytic intervention.24f25 The introduction of geometric and densitometric analysis into clinical practice for the selection of PTCA candi-
Volume Number
115 5, Part 1
PTCA
dates seems therefore justifiable, but the actual meaning of these methods for the evaluation of post-PTCA results remains to be determined. We are indebted to Mr. Alex Crowther for his kind and to Dr. Peter Wilmsburst and Dr. David Thompson comments on statistical analysis.
assistance for their
REFERENCES
1. DeRouen T, Murray JA, Owen W. Variability in the analysis of coronary arteriograms. Circulation 1977;55:324-8. 2 Grondin CM, Dyrda I, Pasternac A, Campeau L, Bourassa MG, Lesperance J. Discrepancies between cineangiographic and postmortem findings in patients with coronary artery disease and recent myocardial revascularisation. Circulation 1974;49:703-8. 3. Harrison DG, White CW, Hiratzka LF, Doty DB, Barnes DH, Eastham CL, Marcus ML. The value of lesion cross-sectional area determined by quantitative coronary arteriography in assessing the physiological significance of proximal left anterior descending coronary arterial stenoses. Circulation 1984; 69:1111-19. 4. Spears JR, Sandor T, Als AV, Malagold M, Markis JE, Grossman W, Serur JR, Paulin S. Computerized image analysis for quantitative measurement of vessel diameter from cineangiograms. 1983;68:453-61. 5. Brown BG, Bolson E, Frimer M, Dodge H. Quantitative coronary angiography. Estimation of dimensions, hemodynamic resistance, and atheroma mass of coronary artery lesion, using the arteriogram and digital computation. Circulation 1977;55:329-37. 6. Nichols AB, Gabrielli CFO, Fenoglio JJ, Esser PD. Quantification of relative coronary arterial stenosis by cinevideodensitometric analysis of coronary arteriograms. Circulation 1984;69:512-22. 7. Wiesel J, Grunwald AM, Tobiasz C, Robin B, Bodenheimer MM. Quantitation of absolute area of a coronary arterial stenosis: experimental validation with a preparation in vivo. Circulation 1986;74:1099-1106. 8. Bove AA, Holmes DR, Owen RM, Bresnahan JF, Reeder GS, Smith HC, Wlietstra RE. Estimation of the effects of angioplasty on coronary artery stenosis using quantitative video angiography. Cathet Cardiovasc Diagn 1985;11:5-16. 9. Meier B, Gruentzig AR, Goebel N, Pyle R, Von Gosslar W, Schlumpf M. Assessment of stenoses in coronary angioplasty. Interand intraobserver variability. Intern J Cardiol 1983;3:159-69. 10. Block PC, Myler RK, Stertzer S, Fallon JT. Morphology after transluminal angioplasty in human beings. N Engl J Med 1981;305:382-8. 11. Holmes DR, Vlietstra RE, Mock MB, Reeder GS, Smith HC, Bove AA, Bresnahan dF, Pichler JM, Schaff HV, Orszulak TA. Angiographic changes produced by percutaneous transluminal coronary angioplasty. Am J Cardiol 1983;51:676-.83.
assessment
by computed
DSA
1187
12. Serruys PW, Reiber JHC, Wijns W, von den Brand M, Koojman CJ, den Katen HJ, Hugenholtz PG. Assessment of percutaneous transluminal coronary angioplasty by quantitative coronary angiography: diameter versus densitometric area measurements. km J-Cardiol 1984;54:482-8. 13. Tobis J. Nacioelu 0. Johnston WD. Qu L. Reese T. Sato D. Roeck W, Monielli S, Henry WL. Vidkodensitometric deter: mination of minimum coronary artery luminal diameter before and after angioplasty. Am J Cardiol 1987;59:38-44. 14 Sanz ML, Mancini J, LeFree MT, Michelson JK, Starling MR, Vogel RA, Top01 EJ. Variability of quantitative digital subtraction coronary angiography before and after percutaneous transluminal coronary angioplasty. Am J Cardiol 1987;60:55-60. 15. Pope DL, Parker DL, Clayton PD, Gustafson DE. Left ventricular border recognition using a dynamic search algorithm. Radiology 1985;155:513-18. 16. Mancini GBJ, Higgins CB. Digital subtraction angiography: a review of cardiac applications. Prog Cardiovasc Dis 1985; 28:111-41. 17. Mancini GBJ, Simon SB, McGillem MJ, LeFree MT, Friedman HZ, Vogel RA. Automated quantitative coronary arteriography: morphologic and physiologic validation in vivo of a rapid digital angiographic method. Circulation 1987;75:45260. 18. Klein LW, Agarwal JB, Rosenberg MC, Stets G, Weintraub WS, Schneider RM, Hermann G, Helfant RH. Assessment of coronary artery stenoses by digital subtrat tion angiography: a pathoanatomic validation. AM HEART J 1987:113:1011-17. 19. Gould KL, Lipscomb K, Hamilton GW. Physiological basis for assessing critical coronary stenosis. Am J Cardiol 1979; 33:87-94. 20. Gould KI, Lipscomb K. Effect of coronary stenosis on coronary flow reserve and resistance. Am J Cardiol 1974; 34:48-55. 21. Wilson RF, Marcus ML, White CW. Prediction of the physiologic significance of coroary arterial lesions by quantitative lesion geometry in patients with limited coronary artery disease. Circulation 1987;75:723-32. 22. Ischinger T, Gruentzig AR, Hollman J, King S, Douglas J, Bradford J. Should coronary arteries with less than 60”~ diameter stenosis be treated by angioplasty? Circulation 1982:66(Suppl. 11)329,1982. 23. Zijlstra F, van Ommeren J, Reiber JHC, Serruys PUT. Does the quantitative assessment of coronary artery dimensions predict the physiological significance of coronary stenosis? Circulation 1987;75:1154-61. “4. Harrison DG, Ferguson DW, Collins SM, Skorton DJ. Ericksen EE, Kioschos JM, Marcus ML, White CW. Rethrombosis after reperfusion with streptokinase: importance of geometry of residual lesions. Circulation 1984;5:991-9. “5. Sheehan FHI Mathey DG, Schafer J, Dodge HT, Bolson EL. Factors that determine recovery of left ventricular function after thrombolysis in patients with acute myocardial infarction. (iirculation 1985;6:1121-8.
Demosthenis Katritsis, Ian C. Cooper, MRCP,
MD, David A. Lythall, MB, BS, Mark H. Anderson, MRCP, and Michael M. Webb-Peploe, FRCP. London, England
Subjective visual assessment of percent stenosis of coronary arteries from routine cineangiograms is subject to large intra- and interobserver variability,’ and a poor correlation between angiographic and postmortem pathologic findings has been demonstrated.* Previous studies3 have also shown that the minimal cross-sectional area of a stenosis rather than its percent diameter reduction is the better determinant of its hemodynamic significance. In order to assess more precisely the functional significance of coronary artery stenoses, computer-aided measurements of the geometric dimensions of the stenosis4*5 and densitometric evaluation of its crosssectional area have been used.4,6* 7 Analysis of coronary angioplasty (PTCA) results presents particular problems. It has been showns,s that in severe lesions visual estimation tends to overstate the severity of the pre-PTCA obstruction. From
the Department
Supported Received Reprint Hospital,
of Cardiology,
by the Cardiac for publication requests: London
Research Feb.
5, 1988;
D. Katritais, MD, SE1 7EH, England.
St. Thomas’s Fund,
accepted Dept.
Hospital.
St. Thomas’s July
Hospital. 1, 1988.
of Cardiology,
St. Thomas’s
PTCA usually causes intimal tearing and gross di$ortion of the angiographic appearance of the artery.lO* l* Important discrepancies between geometric and densitometric measurements of post-angioplasty lumens may therefore occur, and densitometric analysis in a single view has been proposed for the evaluation of PTCA results.‘* In other studies, however, densitometry has not been found significantly different from edge detection,13 and an impressive lack of correlation between densitometric measurements in orthogonal views, both pre- and post-PTCA, has been demonstrated.14 All these reports,i2-14 however, have been based on relatively small numbers of cases. In this study we applied an automated computer program capable of simultaneous geometric and densitometric analysis of digital angiograms (DSA) for evaluation of the results of PTCA. The aims were to: (1) compare these measurements with the visual interpretation of routinly reported 35 mm cineangiograms; (2) test the validity of densitometric analysis in a single angiographic projection for the assessment of pre- and post-angioplasty stenoses; and (3) compare the geometric and densito1181
1182
Kntritsis
et nl.
American
Table
I. Coronary
artery
lesions dilated
by PTCA
No. of patients Total lesions LAD RCA
LAD, left anterior SVBG. saphenous
Fig. 1. Digital subtraction angiograms showing geometric and densitometric calculations before (top panel) and after angioplasty (bottom panel).
metric characteristics cedure.
of the lumens after the pro-
METHODS Imaging system. Angiograms were obtained on a realtime digital image acquisition and processing system (Digitron 2, Siemens AG, Munich, W. Germany) interfaced to a conventional cineangiographic unit (Pandoros Generator/Cardioscop U, Siemens AG, Erlangen, W. Germany). An image module with 20 megabyte semiconductor memory was used for acquisition, and postnrn~nr.,.“:r,- ...ua ,,,*X uui on a videu processor equipped with a 168 megabyte Winchester disk drive. Images were acquired at 8 frames/set in a 512 X 512 matrix with lo-bit depth. Prior to digitization, logarithmic preamplification
November 1988 Heart Journal
63 73
cx
34 15 13
SVBG
11
descending; vein bypass
RCA. graft.
right
coronary
artery;
Cx, circumflex;
compensated for nonlinear x-ray absorption resulting from different object density and thickness (LambertBeer law). Phasic mask subtraction was used, i.e., diastolic mask frames were used for subtraction from diastolic images and systolic masks were used for systolic images. Cineangiograms at the same views were acquired at 25 frames/set. Cineangiograms were routinely reported by the cardiologist who had performed the angioplasty and who was unaware of the DSA measurements. Analysis of the DSA images was performed on those views (one to four) in which the pre-PTCA lesion was clearly seen and the same views were used for the postPTCA analysis. In this way 58 lesions were studied in two to four projections and 15 lesions were studied in one projection only. Individual images were selected from both the pre- and the post-PTCA recordings and were subsequently stored on the Winchester disk. An image was selected for analysis that met the following criteria: (1) The image was selected from such a phase of the contrast injection that both the lesion and the adjacent normal segments of the artery were equally opacified. (2) All segments had to be displayed without foreshortening or overshadowing by side branches or other anatomic structures. (3) The compared images were selected at identical times of the cardiac cycle (late diastole). The same views and similar images were selected for the pre- and post-angioplasty recordings. For the quantitative studies, each image was recalled on the monitor and was submitted to geometric and densitometric analysis performed by a computer program based on a borderdetecting dynamic search algorithm.15 Edge detection program. This program is based on a modular, fully automated border detection algorithm developed by Pope, Parker, et al. and described in detail elsewhere. I5 Having displayed the preselected image on the monitor, the operator, with a light-pen cursor, defines the search area the program will use to find the vessel edges and the complete vessel length to be analyzed. Geometric cross-sectional area at each point along the centerline is obtained by assuming circular cross-sections and densitometric cross-sectional area is obtained by integrating the density across the perpendicular profile from edge to edge. The density values are background corrected by subtracting an average background value determined by the densities beyond the edges of the vessel. Two pairs of cursors representing the largest and smallest values are then automatically placed over the
Volume Number
Table
116 5, Part 1
II. Comparison
PTCA
of computer
estimations
on subtracted
and unsubtracted
assessment
Densitometric
area % stenosis
Values are expressed as mean + SEM. r, correlation coefficient; Sub, subtracted;
1183
Post-PTCA Unsub
Sub
Geometric area % stenosis
DSA
images
Pre-PTCA
Diameter % stenosis
by computed
Sub
58.3 + 3.5 59.7 * 3.3 (r = 0.90, p < 0.001) 81.4 + 2.8 82.5 f 2.9 (r = 0.88, p < 0.001) 80.0 f 3.1 80.1 + 3.6 (r = 0.89, p < 0.001)
Unsub
36.3 k 6.5 36.7 (r = 0.97, p < 0.001) 55.1 -+ 9.2 55.4 0. = 0.97, p < 0.001) 42.6 I+ 9.7 43.9 (r = 0.96, p < 0.001)
k 6.8 5 9.5 + 9.4
Unsub, unsubtracted.
vessel and their position is further adjusted by the operator. Percent diameter stenosis, percent geometric crosssectional area stenosis, and percent densitometric crosssectional area stenosis are then automatically displayed (Fig. 1). The stenotic and reference areas are reproduced on the post-PTCA images. For statistical analysis, the measurements obtained in all the angiographic projections studied were averaged for each method and every lesion. Calibration of the digital images can be achieved by using the inside diameter of the catheter guiding system as reference. Strictly, absolute calibration of the system is unnecessary when comparing relative percentage stenoses before and after PTCA. Such comparison of relative stenoses also avoids potential errors from pincushion distortion, causing selective magnification of the peripherally located guiding system. Subtracted and unsubtracted images. Automated border detection was applied to 13 projections of 6 lesions in both subtracted and image (unsubtracted) modes and the results were compared. In two instances, border detection on unsubtracted images was not possible; the lesion was overshadowed by either the pulmonary artery (pacing) catheter or, in the case of a caudal view, by the spine. In the rest of the projections, geometric and densitometric stenoses derived from the two image modalities were similar with a very good correlation (presented in the Results section). Consequently, only subtracted images were selected for further analysis. Statistical analysis. Student’s two-tailed paired t test and linear regression analysis were used for the comparison of stenoses as assessed by different methods. In addition, the mean and the standard deviation of their differences were considered. Because some data were not normally distributed (i.e., visual reports), nonparametric statistical tests were additionally employed (Wilcoxon signed-rank test and Spearman’s rank correlation). The results, however, did not differ significantly from those of parametric analysis. Thus parametric results only are presented in the text. The coefficient of variation, as a percentage, was used to assess the value of taking multiple views for densitometric analysis. Patients. Sixty-three patients who had undergone successful PTCA were studied. PTCA was performed on 73 lesions that were thought to be responsible for exercise-
induced myocardial ischemia demonstrated by thallium201 scintigraphy. The following categories were not included in the study. (1) Cases with obvious postangioplasty dissection interfering with the program’s edge detection. (2) Patients who had previously undergone coronary artery bypass grafts and had angioplasty of their native vessels. The native coronaries were usually distorted due to the grafts, so it was difficult to define the reference (normal) part of the vessel. (3) Lesions that could not. be displayed without their being overshadowed by side branches, whatever projection was used. The sites of the studied lesions are summarized in Table I. All lesions were evaluated by means of visual assessment and combined geometric and densitometric quantitation by the automated computer program. RESULTS Subtracted
statistically
vs unsubtracted images. There was no significant difference in diameter, geo-
metric area, or densitometric area between subtracted and unsubtracted
percent
stenoses
images, either pre- or post-PTCA. A very good correlation existed between the two image modalities, whatever kind of measurement was considered (Table II). Pre-PTCA VI post-PTCA results. The percent stenoses as evaluated by the various methods both before and after PTCA are shown in Table III. There was a significant improvement in luminal diameter after PTCA (p < O.OOl), whatever method was used. Visual
reports
vs computer
diameter
measurements.
The visual reports of cineangiograms significantly overestimated the pre-PTCA diameter percent stenosis and underestimated the post-PTCA residual stenosis in comparison with the computer program (p < 0.001 in both cases). Fig. 2 and Table IV show the important disagreement between the two methods and the virtual lack of any correlation between them.
Before PTCA, a total of 43 out of the 73 lesions dilated (i.e., 59%) were estimated to represent less than 60% diameter stenosis by the computer. Sev-
November
I 184
Katritsis
et al.
American
Table III. Comparison of pre- and post-PTCA estimated visually and by the computer Pre-PTCA
Values
are expressed
as mean
Pre-PTCA
12.7 ” 1.7
55.3 i 1.1
23.2 i 1.8
78.8
+ 1.1
37.9 + 2.7
79.0
+ 1.2
36.9
r P
r, correlation
P
Mean of differences SD of differences r, correlation
coefficient;
Post-PTCA 0.48
0.04 0.76 26.3
-Co.001
-10.5 15.0
13.8 SD, standard
deviation.
enteen lesions (i.e., 23%) were estimated to be less than 50%. All 76 lesions were reported to be more than 60% stenotic before PTCA on visual estimation. Geometric vs densitometric computer measurements. There was no statistically significant difference between geometric and densitometric area stenoses, either pre-PTCA or post-PTCA. The two image modalities yielded variables that were well correlated pre-PTCA (r = 0.82, p < O.OOl), whereas there was a modest correlation post-PTCA (r = 0.71, p < 0.001) (Fig. 3 and Table V). However, analysis of their differences indicated important discrepancies post-PTCA (standard deviation of the differences = 18.6) that did not exist pre-PTCA (standard deviation of the differences = 6.1). Multiple vs single projections; metric evaluation. For those
densitometric
and geo-
58 lesions studied in more than one projection, the coefficient of variation was calculated for the degree of densitometric area stenosis assessed from multiple views of each individual lesion. The mean of the coefficients of variation for all lesions was 8% pre-PTCA and 69 % post-PTCA. This indicates a wide variation of den-
oitr\mntr:n
-7,17-n“c-uIY
~$t~i&
f;cm
the
same
coefficient;
SD, standard
Post-PTCA 0.71
18.6
deviation.
k 3.0
Table IV. Correlation between visual reports and computer measurements of percent stenoses of luminal diameter
r
0.82
Mean of differences SD of differences
i SEM.
Pre-PTCA
1988 Journal
Table V. Correlation between geometric and densitometric evaluations of percent of area stenosed
Post-PTCA
81.6 i- 1.2
PC, Diameter stenosis (visual report) rC1 Diameter stenosis (computer program) “;# Geometric area stenosis (computer program) 5, Densitometric area stenosis (computer program)
stenoses
Heart
lesion
in
different gngiographic projections after PTCA, but not before. In the case of geometric area percent
stenosis, the mean of the coefficient of variation for all lesions was 7% pre-PTCA and 25% post-PTCA. Therefore post-PTCA, geometric areas do not show the same degree of variation demonstrated by densitometry. DISCUSSION
The value of quantitative methods in assessing the physiologic significance of coronary artery disease and its response to therapeutic interventions has been well documented.**9s’3*14 Geometric methods assume either a circular (for one-plane views) or elliptical (for biplane views) cross-sectional shape of the lesion. To avoid geometric assumptions that do not characterize asymmetric lesions, densitometric techniques have made use of the concept that in an ideal setting the density across an artery in a radiographic image is proportional to the crosssectional area at that point.‘j Several studies6*7 have proposed that even a single radiographic projection could be sufficient for the estimation of the degree of stenosis. However, densitometric evaluation is based on several premises.6 First, it is assumed that radiographic magnification and exposure are uniform. Second, the distribution of the dye must be equal in both normal and stenotic arterial segments. In clinical practice this is not always the case. Digital subtraction angiography. Since almost all quantitative methods require digitization of conventional cineangiograms, on-line digital acquisition of coronary arteriograms is ideal for such applications. Cumbersome digitization of cineangiograms is avoided and the need for high frame rate angiographic views is reduced. Suppression of distracting background anatomy and enhancement of the contrast level can significantly improve the yield of diagnostic information. Disadvantages of subtraction angiography include inferior spatial and temporal resolution compared with cineangiograms.16 Furthermore, misregistration artifacts caused by respiratory movement remain one of the most common causes of image degradation when conventional
Volume
116
Number
5, Part 1
PTCA assessment Pre-PTCA
by computed DSA
Post-PTCA
r = 0.04
r = 0.48
T.5 90.P 0” c 80-
n
d . f 70fi; E .m 60 cl El 2 soE 9
1185
q q n
II q q q
. 40m 30:.
, 30
.
,
40
.
50
,
.
60
Visual Report
,
q ,
,
70
.
,
80
Diameter
,
90
Stenosis
(
-20 1,
100
-20
(%)
0
20
Visual Report
40
60
Diameter
80
Stenosis
100
(%)
Fig. 2. Relationship between visual estimation and computer measurements of percent stenosesof luminal diameter.
Pre-PTCA 100
y = 4.76 + 0.94x
r = 0.82
70-
$
40-
$
zo-
.s
Oi? . .s2 -20 E . 8 -4o-
60-
$ (ij
r = 0.71
‘;i 80a. i 60e .
-2 $
y = 7.35 + 0.78x
100 -I
1
7 5 go.I? 0” 5 so(I) zi 2
Post-PTCA
50-
q
q 40
. 40
I
.
50
Geometric
I 60
I
, 70
.
I 80
Area Stenosis
.
, 90
.
, 100
(“I&)
-60
. -60
, -40
.
, -20
Geometric
.
, 0
.
, 20
.
, 40
Area Stenosis
.
, 60
.
, 80
.
( 100
(%)
Fig. 3. Linear regression analysis of geometric and desitometic evaluations of percent of area stenosed.
first-order subtraction is employed. Newer methods such as recursive filtering and “second-order” techniques such as hybrid subtraction should eliminate these problems.16 Recently, Mancini et al.” reported the successful evaluation of a similar DSA program by studying well-defined lesions created by means of cylinders of known diameter placed over canine coronary arteries. This technique allowed accurate luminal diameter measurements, and direct comparison of digital and tine images showed some deterioration in both accuracy and reproducibility when films were used.
Tn vit,ro evaluation of a similar method on human cadaver coronary arteries” showed very good correlation of the geometric and densitometric results with histopathologic measurements. These results indicate that on-line digital images can provide greater morphologic precision than routinely processed cineangiograms. Assessment of PTCA results. Analysis of the results of PTCA presents particular problems. Meier et a1.,g found that the intraobserver coefficient of variation was significantly reduced when the mean estimations from three projections instead of the projec-
November
1166
Katritsis
et al.
tion showing the most severe stenosis were used. Bove et al.* demonstrated that visual assessment overestimated the percent diameter stenosis of severe lesions. Serruys et a1.,12in a study of 18 cases by quantitative methods on conventional angiograms, found a good agreement between the densitometric precent area stenosis and circular percent area stenosis before PTCA. After PTCA, important discrepancies were observed between the two types of evaluation. These investigators suggested that the mechanical disruption of the internal artery wall caused by the procedure leads to eccentric, asymmetric morphologic changes that cannot be accurately assessed from the detected contour of the vessel when a single angiographic view is used. They therefore proposed the use of densitometry as essential for that purpose. However, in a recent digital evaluation of angiographic study, l3 densitometric both the pre- and post-PTCA lumens was not significantly different from the edge detection method, and the correlation between measurements in orthogonal projections was similar for both techniques either before or after PTCA. Sanz et a1.,14in a study of 13 cases with a digital method similar to our own, showed a lack of correlation between densitometric measurements in orthogonal views both before and after PTCA. This study represents the first extensive clinical application of an automated computer-aided technique for digital image quantitation before and after PTCA. It allowed rapid quantitative assessment of coronary artery stenoses both before and after the intervention. The results support the evidence for inaccuracy of visual assessment of coronary stenosis both pre- and post-PTCA. In accordance with previous reports,s we have found that this overestimation is worse with more severe stenoses (i.e., pre-PTCA). For example, the computer found 23 % of all lesions to have a stenosis of less than 50% of the lumen diameter. Previous reports’g’20 have shown that reduction of the arterial lumen causes significant alteration in resting flow only when the stenosis is greater than 85 % , and that lumen stenosis may limit significantly maximal flow response to hyperemic stimuli when it is greater than 45%. Recently, Doppler catheter studieszl for the assessment of coronary flow reserve demonstrated that maximal hyperemic coronary flow is usually impaired by stenoses affecting more than 70% of the luminal area or by those producing more than 50% diameter obstruction. Furthermore, it has been shownz2 that U~IgLpiasLy or stenoses of less than 60% carries a substantial risk of complications and is better avoided. The relevance of these findings to the
American
1988
Heart Journal
selection of PTCA candidates is obvious. Overestimation of diameters of less than 1 mm is a welldescribed feature of automated quantification techniques.16 Phenomena such as motion blurring, limited spatial resolution, and geometric magnification all promote overestimation of diameters at this level. Mancini et a1.,17however, accurately measured diameters as low as 0.8 mm, and pre-PTCA diameters averaging 1.0 mm8 or 1.28 mm12 are the general rule. Such overestimation would presumably affect very tight lesions and not the 50% to 60% stenoses where our visual report and the computer disagreed. The program’s geometric (cross-sectional area) assessment of the pre-PTCA stenoses correlated well with its densitometry derived values and the discrepancies between the two methods do not seem to be of clinical importance. In the post-PTCA assessment, however, discrepancies were noted between the two methods, in accordance with the findings of previous reports.12 In contrast to the report of Sanz et a1.,14 before PTCA both densitometric and geometric analysis demonstrated low mean coefficients of variation in the projections best demonstrating the lesion. PostPTCA, however, the densitometric variation in the same views was much larger than the geometric variation, suggesting that one projection is not enough for densitometric analysis. It has been assumed that following angioplasty, distortion of the vessel lumen in most cases makes any accurate morphologic description difficult. The fact that this degree of variation does not exist in geometric evaluation post-PTCA suggests that the densitometric discrepancies between views cannot be ascribed to the irregular anatomy of the postangioplasty lumen alone. Similarly, inaccuracies of the densitometric method itself do not appear to provide an adequate explanation, since this variation did not exist pre-PTCA. It may be that the distortion of the vessel interferes with the mixing of the radiographic contrast medium and blood, hence invalidating any assumptions about dye distribution. Quantitative methods that use edge dection programs seem to be accurate and reproducible,14 and may have both diagnostic and prognostic significance. Stenotic diameters or cross-sectional areas acquired in this way have been found to indicate with reasonable accuracy the consequent reduction in coronary flow capacity21,23 and to predict reocclusion rates or left ventricular function recovery following thrombolytic intervention.24f25 The introduction of geometric and densitometric analysis into clinical practice for the selection of PTCA candi-
Volume Number
115 5, Part 1
PTCA
dates seems therefore justifiable, but the actual meaning of these methods for the evaluation of post-PTCA results remains to be determined. We are indebted to Mr. Alex Crowther for his kind and to Dr. Peter Wilmsburst and Dr. David Thompson comments on statistical analysis.
assistance for their
REFERENCES
1. DeRouen T, Murray JA, Owen W. Variability in the analysis of coronary arteriograms. Circulation 1977;55:324-8. 2 Grondin CM, Dyrda I, Pasternac A, Campeau L, Bourassa MG, Lesperance J. Discrepancies between cineangiographic and postmortem findings in patients with coronary artery disease and recent myocardial revascularisation. Circulation 1974;49:703-8. 3. Harrison DG, White CW, Hiratzka LF, Doty DB, Barnes DH, Eastham CL, Marcus ML. The value of lesion cross-sectional area determined by quantitative coronary arteriography in assessing the physiological significance of proximal left anterior descending coronary arterial stenoses. Circulation 1984; 69:1111-19. 4. Spears JR, Sandor T, Als AV, Malagold M, Markis JE, Grossman W, Serur JR, Paulin S. Computerized image analysis for quantitative measurement of vessel diameter from cineangiograms. 1983;68:453-61. 5. Brown BG, Bolson E, Frimer M, Dodge H. Quantitative coronary angiography. Estimation of dimensions, hemodynamic resistance, and atheroma mass of coronary artery lesion, using the arteriogram and digital computation. Circulation 1977;55:329-37. 6. Nichols AB, Gabrielli CFO, Fenoglio JJ, Esser PD. Quantification of relative coronary arterial stenosis by cinevideodensitometric analysis of coronary arteriograms. Circulation 1984;69:512-22. 7. Wiesel J, Grunwald AM, Tobiasz C, Robin B, Bodenheimer MM. Quantitation of absolute area of a coronary arterial stenosis: experimental validation with a preparation in vivo. Circulation 1986;74:1099-1106. 8. Bove AA, Holmes DR, Owen RM, Bresnahan JF, Reeder GS, Smith HC, Wlietstra RE. Estimation of the effects of angioplasty on coronary artery stenosis using quantitative video angiography. Cathet Cardiovasc Diagn 1985;11:5-16. 9. Meier B, Gruentzig AR, Goebel N, Pyle R, Von Gosslar W, Schlumpf M. Assessment of stenoses in coronary angioplasty. Interand intraobserver variability. Intern J Cardiol 1983;3:159-69. 10. Block PC, Myler RK, Stertzer S, Fallon JT. Morphology after transluminal angioplasty in human beings. N Engl J Med 1981;305:382-8. 11. Holmes DR, Vlietstra RE, Mock MB, Reeder GS, Smith HC, Bove AA, Bresnahan dF, Pichler JM, Schaff HV, Orszulak TA. Angiographic changes produced by percutaneous transluminal coronary angioplasty. Am J Cardiol 1983;51:676-.83.
assessment
by computed
DSA
1187
12. Serruys PW, Reiber JHC, Wijns W, von den Brand M, Koojman CJ, den Katen HJ, Hugenholtz PG. Assessment of percutaneous transluminal coronary angioplasty by quantitative coronary angiography: diameter versus densitometric area measurements. km J-Cardiol 1984;54:482-8. 13. Tobis J. Nacioelu 0. Johnston WD. Qu L. Reese T. Sato D. Roeck W, Monielli S, Henry WL. Vidkodensitometric deter: mination of minimum coronary artery luminal diameter before and after angioplasty. Am J Cardiol 1987;59:38-44. 14 Sanz ML, Mancini J, LeFree MT, Michelson JK, Starling MR, Vogel RA, Top01 EJ. Variability of quantitative digital subtraction coronary angiography before and after percutaneous transluminal coronary angioplasty. Am J Cardiol 1987;60:55-60. 15. Pope DL, Parker DL, Clayton PD, Gustafson DE. Left ventricular border recognition using a dynamic search algorithm. Radiology 1985;155:513-18. 16. Mancini GBJ, Higgins CB. Digital subtraction angiography: a review of cardiac applications. Prog Cardiovasc Dis 1985; 28:111-41. 17. Mancini GBJ, Simon SB, McGillem MJ, LeFree MT, Friedman HZ, Vogel RA. Automated quantitative coronary arteriography: morphologic and physiologic validation in vivo of a rapid digital angiographic method. Circulation 1987;75:45260. 18. Klein LW, Agarwal JB, Rosenberg MC, Stets G, Weintraub WS, Schneider RM, Hermann G, Helfant RH. Assessment of coronary artery stenoses by digital subtrat tion angiography: a pathoanatomic validation. AM HEART J 1987:113:1011-17. 19. Gould KL, Lipscomb K, Hamilton GW. Physiological basis for assessing critical coronary stenosis. Am J Cardiol 1979; 33:87-94. 20. Gould KI, Lipscomb K. Effect of coronary stenosis on coronary flow reserve and resistance. Am J Cardiol 1974; 34:48-55. 21. Wilson RF, Marcus ML, White CW. Prediction of the physiologic significance of coroary arterial lesions by quantitative lesion geometry in patients with limited coronary artery disease. Circulation 1987;75:723-32. 22. Ischinger T, Gruentzig AR, Hollman J, King S, Douglas J, Bradford J. Should coronary arteries with less than 60”~ diameter stenosis be treated by angioplasty? Circulation 1982:66(Suppl. 11)329,1982. 23. Zijlstra F, van Ommeren J, Reiber JHC, Serruys PUT. Does the quantitative assessment of coronary artery dimensions predict the physiological significance of coronary stenosis? Circulation 1987;75:1154-61. “4. Harrison DG, Ferguson DW, Collins SM, Skorton DJ. Ericksen EE, Kioschos JM, Marcus ML, White CW. Rethrombosis after reperfusion with streptokinase: importance of geometry of residual lesions. Circulation 1984;5:991-9. “5. Sheehan FHI Mathey DG, Schafer J, Dodge HT, Bolson EL. Factors that determine recovery of left ventricular function after thrombolysis in patients with acute myocardial infarction. (iirculation 1985;6:1121-8.