Author + information
- Received December 3, 1996
- Revision received October 23, 1997
- Accepted December 18, 1997
- Published online March 15, 1998.
- Giorgio Tommasini, MDABC,* (, )
- Alberto Camerini, MDA,
- Antonio Gatti, MDC,
- Giorgio Derchi, MDC,
- Alessandro BruzzoneB and
- Carlo Vecchio, MDC
- ↵*Dr. Giorgio Tommasini, Divisione di Cardiologia, Ospedale di Treviglio, Piazza Ospedale 1, 24047 Treviglio (Bg), Italy.
Objectives. We describe a new imaging technique for coronary angiography.
Background. The conventional approach to coronary angiography exploits static perspective imaging over multiple cardiac cycles, using a limited number of empirically selected views. This approach entails both lack and redundancy of information and may result in suboptimal visualization of the individual lesion, contributing to diagnostic inaccuracy.
Methods. We developed a new imaging technique exploiting dynamic perspective, obtained by transverse 180° rotation of the Carm of a conventional angiographic unit during standard selective coronary opacification and filming. This technique yields a picture of the coronary tree isocentrically rotating around the longitudinal axis and conveying complete three-dimensional information.
Results. A complete diagnostic run for both coronary arteries, including two 25° cranial and two 25° caudal scans is accomplished with a total cine time of 16 s and 45 ml of contrast medium, about half of that required by conventional angiography. In a series of 129 consecutive patients studied by both the conventional and the new technique with quantitative measurements of the severity of the stenoses, the final diagnosis was identical in 65. In no case was a stenosis detected only by the conventional approach. However, in 31 patients the new technique permitted identification of 34 critical stenoses (79 ± 8% [mean ± SD]) either underestimated (61 ± 3% n = 24, p < 0.001) or undetected (21 ± 22%, n = 10, p < 0.001) in the standard projections. In a further 28 cases, 33 subcritical lesions (60 ± 5%) were visualized in the rotational images but were insignificant (24 ± 22% p < 0.001) in the standard projections. In five additional patients, distinct laminar plaques were clearly visualized only by the panoramic approach.
Conclusions. This new technique can be easily implemented on conventional angiographic equipment at no additional cost. It provides complete, operator-independent exploitation of the angiographic information, resulting in enhanced diagnostic accuracy.
Despite technologic advancement, coronary angiography has limited diagnostic accuracy, as demonstrated in comparative studies by pathologic analysis [1–5]and more recently by intracoronary ultrasound [6–8]and angioscopy [9–11]. There is increasing evidence that plaque morphology and the extent of lumen narrowing cannot be reliably predicted by angiographic appearance, particularly in the presence of eccentric or complicated lesions [2, 3, 5]. Some diagnostic inaccuracy could be ascribed to use of suboptimal projections or, more generally, to incomplete exploitation of the angiographic information. This hypothesis is suggested by the following considerations: 1) Only a single, specific projection (i.e., perpendicular to the outline of maximal narrowing) will permit reliable quantification of the severity of a stenosis in a single frame; 2) each transaxial view conveys independent, additional information and may be essential to elucidate lesion morphology; 3) use of all transaxial projections is a requisite for exact three-dimensional reconstruction.
In contrast to these requirements, coronary angiography actually consists of a limited number of predetermined projections, individually adjusted by the operator according to the presumed geometry and orientation of the stenoses. The choice of the views is thus in part arbitrary and partly follows a trial and error process that should be applied to each lesion to get optimal visualization. However, because the overall number of projections is limited by time, safety and cost, the usual compromise is to obtain four to seven projections for the left and two to four for the right coronary artery. The resulting gap between adjacent projections, and thus the potential deviation from the optimal angle of observation, will range from 30° to >90° when only two projections are used. This gap can lead to serious underestimation of the severity of the stenosis and of its length (see later). Besides being incomplete, this information entails considerable redundancy because each projection includes several cardiac cycles, yielding a series of highly intercorrelated images. From all these limitations, it appears clear that the conventional approach is not optimal.
To overcome these limitations we developed a new approach that uses a dynamic rather than a fixed perspective, obtained by transverse 180° rotation of the Carm of a conventional angiographic unit, accomplished manually in 4 s during standard selective coronary opacification and filming. The basic principle was suggested by Spears et al. several years ago. However, to our knowledge, this technique has never been applied to the imaging of the coronary arteries.
The present report describes our experience with the aforementioned approach, compared with the standard technique, in patients with coronary artery disease.
1.1 Theoretical Background
To evaluate the influence of foreshortening on the apparent length and severity of the stenosis we used a simple model of a concentric stenosis, based on a narrow tube (the stenosis) interposed between two larger tubes of equal radius (the normal segments) aligned on the same axis. This model and the corresponding silhouette from three different perspectives are depicted in Fig. 1. The correct perspective, which avoids foreshortening, is perpendicular to the axis of the tubes. Deviation from this perspective leads to progressive underestimation of the length and, above a certain threshold, of the severity of the stenosis.
In geometric terms, the stenosis is described by the percent diameter reduction S= (1 − d/D) × 100, where d and D are the diameters of the stenotic and the normal segments, respectively, and by the ratio ϵ between the length of the stenotic segment L and the diameter of the normal segment D.
The perspective error α is defined as the angle between the actual and the correct perspective. When α becomes greater than a limit angle α lim, the apparent diameter of the stenotic segment starts to increase, leading to progressive underestimation of stenosis severity.
For this model of a concentric stenosis, α lim is calculated as follows: the resulting error e, defined as the difference between apparent S′ and true percent stenosis S, corresponds to Thus, the degree of underestimation is directly related to perspective error and inversely related to the relative length of the stenosis. Fig. 2shows the error e as a function of α for three stenoses with S = 70% and ϵ = 0.25, 0.5 and 1.0, respectively.
1.2 Coronary Angiography
At present, no commercially available cardioangiographic unit can accomplish a 180° automated rotation of the Carm around the patient at a speed compatible with the brief duration of a standard coronary injection. Accordingly, the present study was carried out using a system (Philips Poly C2) that allows for manual rotation of the Carm.
The technique we adopted is as follows. The position of the patient was initially adjusted under fluoroscopy so that the heart lay approximately isocentric to the Carm. The image intensifier was 25° cranially or caudally tilted, then positioned 90° right lateral, close to the thoracic wall, and the Carm was manually rotated from the right to the left side during coronary injection and filming (Fig. 3).
Selective coronary opacification was carried out with 7 to 15 ml of lopamidol, manually injected into the coronary ostium while the patient maintained a deep inspiration. Rotation of the Carm started immediately after complete filling of the coronary artery.
A complete 180° scan was accomplished within ∼4 s. A specifically designed mechanical stop was used to prevent displacement of the image intensifier by centrifugal forces during high speed rotation.
All images were recorded on a Grundig 1-in. video tape recorder and filmed at 50 frames/s on an Arrhytecno camera, using an Agfa Scopix RP1C standard cine film. To cope with rapid radiodensity variations occurring during rotation (i.e., over the spine or diaphragm), X-ray exposure was set to automatic mode, using a sensing field of 40%. Nominal exposure was 14 μR/frame; exposure time was 3.6 to 5.0 ms. A 7.5-in. image intensifier field and a focal spot of 0.8 mm were usually employed. Under these conditions >90% of the response of automatic exposure occurs within one frame. This is sufficient to overcome partial volume effect and preserve image quality during rotation.
One hundred twenty-nine consecutive patients with coronary artery disease undergoing diagnostic coronary angiography were studied by both the conventional and the new approach. Written informed consent to the whole procedure was obtained from all patients, and the study protocol was approved by the institutional ethics committee.
Standard projections, basically 10° right anterior oblique (RAO) or 10° left anterior oblique (LAO) with either cranial or caudal tilt, 30° RAO with both cranial and caudal tilt, 45° caudal LAO, 60° cranial LAO and 90° left lateral (LL), were routinely used for the left coronary artery. For the right coronary artery, 45° cranial LAO and 30° RAO projections with both cranial and caudal tilt were routinely used. The degree of tilting was individually adjusted (up to 45°) for optimal visualization. Additional views were also used to elucidate coronary anatomy when requested by the operator. An average of 6.4 and 3.3 projections were used for the left and the right coronary artery, respectively.
After completion of the conventional diagnostic study, two rotational scans with fixed 25° cranial and 25° caudal angulation were performed for each coronary artery.
All films were analyzed by three experienced angiographers using a Tagarno 35CX projector interfaced through a T35W cinevideo camera to a Matrox Image 640 digital processor. In the presence of a discrete stenosis, as identified by visual inspection of the entire film, three consecutive frames consensually judged to show maximal stenosis severity were selected within both the standard and the rotational images. Each frame was digitized at a 640 × 512 pixel resolution; reduction of electronic noise from the TV camera was obtained by averaging 32 consecutive cinevideo scans of the same frame [13, 14]. The severity of each stenosis was assessed by semiautomatic quantitative analysis, using automatic edge detection and diameter measurement on operator-defined crosslines at stenosis and reference locations . The severity of the stenosis was expressed as the mean percent diameter reduction in the three selected frames. Undetectable lesions were quantified as 0% stenosis. All stenoses were classified as either critical(>70%), subcritical(≥50%) or insignificant(<50%).
1.3 Statistical Methods
Results are expressed as mean value ± SD. The Student paired ttest was used to compare the severity of each stenosis by the conventional and the new approach. A p value <0.05 was considered significant.
No complications occurred as a result of the experimental part of study. In all cases, adequate films could be obtained.
2.1 Image Quality
No significant deterioration of image quality due to rotation of the Carm was observed when exposure time was ≤5.0 ms.
Sequential visualization from a perspective continuously variable along an arc of 180° yields an image of the coronary tree isocentrically rotating around the longitudinal axis, as if viewed by an observer evenly moving around the heart (Fig. 4). This allows intuitive reconstruction of three-dimensional anatomy, easy identification of each ramus and appreciation of its territory of distribution and collateral circulation.
2.3 Stenoses Visualization
In 65 patients (50%), an identical diagnosis (i.e., classification of each stenosis in the same category) was made by the conventional and the new technique. In no case was a stenosis detected by the conventional approach only. However, in 31 patients (24%), the panoramic technique permitted recognition of 34 critical stenoses (>70%, mean 79 ± 8%), classified as either subcritical (<70%, mean 61 ± 3%, n = 24, p < 0.001) or insignificant (<50%, mean 21 ± 22%, n = 10, p < 0.001) by the conventional approach. Critical lesions were detected in all segments of the coronary tree, as reported in Table 1. An example of a severe stenosis of the left main coronary artery, underestimated by the standard projections, is shown in Fig. 5.
In a further 28 patients, 33 subcritical lesions (between 50% and 70%, mean 60 ± 5%) were identified by the panoramic technique but were classified as insignificant (<50%, mean 24 ± 22%, p < 0.001) in the standard projections.
In five additional patients, distinct laminar plaques protruding into the lumen were also clearly recognizable from a very narrow perspective in the rotational images only.
3.1 Previous Studies
Over the past three decades, coronary angiography has been subjected to a steady process of refinement, involving several aspects of the technique, to improve both safety and diagnostic accuracy. Although both aims have been approached, major technologic advancements concerning X-ray generators, image intensifiers, TV chains and digital processing, leading to excellent image quality, have been shown to be insufficient to fulfill the expectations insofar as diagnostic accuracy is concerned.
The limitations of coronary angiography for predicting pathologic findings have been repeatedly reported. In particular, the potential for underestimation of the severity of the stenosis was highlighted by Vlodaver et al. , Grondin et al. , Schwartz et al. , Isner et al. and Arnett et al. , who reported that this problem occurred in a proportion of patients ranging from 15% to 37% and indicated inadequate projections, especially in the presence of eccentric lesions as the most likely reasons for the discrepancy. Diagnostic inaccuracy was also reported by MacPherson et al. , in a comparative in vivo work by epicardial high frequency echocardiography, and more recently by Nissen et al. , De Scheerder et al. and Mintz et al. by intravascular ultrasound and by Ramee et al. , Mizuno et al. and Ventura et al. by angioscopy. All these studies show that prediction of plaque morphology, precise evaluation of the extent of lumen narrowing and occasionally even mere detection of its presence can be an elusive task for coronary angiography. This is of concern, especially in view of the growing interest in quantification of stenoses in terms of absolute accuracy, reproducibility and independence of human interaction, up to the development of deterministic three-dimensional models able to predict the hemodynamic consequences of a stenosis [18, 19]. Despite sophistication, these methods are likely to perform poorly if applied to images not truly representative of the geometry of the stenosis.
These pitfalls have been explained by luminology . However, incomplete exploitation of the angiographic information could also contribute to diagnostic inaccuracy. Although the theoretic necessity of imaging each coronary segment as perpendicularly and circumferentially as possible was emphasized several years ago, no practical application fulfilling these requirements has been developed so far. The only significant advancement was the introduction of cranially and caudally oriented views. By this approach, it could be demonstrated that the use of additional perspectives with longitudinal angulation permits identification of significant stenoses not otherwise detectable [21, 22]. However, the choice of the degree of both transaxial and craniocaudal angulation as well as the overall number of projections remains a subjective matter, which cannot guarantee optimal visualization of the individual lesion. More objective, computer-assisted techniques, attempting to identify the optimal angle of observation from biplane views are currently under evaluation [23, 24]. In any case, the conventional multiview approach retains the properties of an ill-posed problem with nonunivocal solutions and is unable to convey complete three-dimensional information.
3.2 Advantages of the Rotational Technique
The rotational technique can overcome some of the aforementioned limitations. It offers the following advantages:
1. Visualization of every lesion from all perspectives in the transverse plane in a single run, yielding maximal information with no redundancy.
2. Panoramic representation of the coronary tree as a rotating image conveying complete three-dimensional information. This gives a better insight into the coronary tree and collateral circulation and permits a more accurate reconstruction of complex anatomies.
3. Standardized, operator-independent approach. The panoramic technique does not rely on presumptive hypotheses on actual anatomy and does not involve an empirical choice of the most suitable projection to visualize a specific lesion. All information is obtained by two rotational scans with fixed 25° cranial and caudal angulation. Hence, the new approach is standardized and basically operator independent, which is expected to improve reproducibility.
4. Improved diagnostic accuracy. In this study, comparative analysis of the results was carried out by using the conventional classification of stenoses based on percent diameter reduction. This approach was adopted since it corresponds to the standard by which most laboratories interpret the severity of a stenosis for clinical purpose. By this standard, the panoramic approach shows improved diagnostic accuracy. In accordance with the theory, the use of all transaxial projections allowed identification of a substantial number of critical stenoses which were either underestimated or undetected by the conventional technique. Also, several subcritical but significant (>50% and <70%) lesions were only detected by the new technique. Overall, a diagnostic advantage could be demonstrated in 50% of patients. The main sources of this additional information were found to be a very steep cranial RAO, the cranial LL and various shallow cranial and caudal LAO views. However, it is unlikely that by simply adding these views to the standard approach to coronary angiography one would obtain the same information because many lesions appeared critical only from a very narrow perspective. This also applied to visualization of laminar lesions, always overlooked in the standard projections. Improved appreciation of lesion morphology actually emerged as a consistent advantage of the new technique.
5. Use of less contrast medium and shorter cine runs. A complete rotational study comprises four scans, ∼45 ml of contrast and 16 s of cine, approximately half required by conventional coronary angiography.
3.3 Study Limitations
A potential limitation of the rotational approach is that it does not allow for diastolic visualization of each stenosis from all perspectives. The choice of end-diastole as the reference phase of the cardiac cycle for quantification of coronary stenoses is theoretically sound but has never been substantiated by specific studies. A recent report by Seltzer et al. focused on the influence of the phase on quantification of the stenoses, and concluded that end-diastolic measurements were less dependent on frame selection within a single cycle. However, random sampling was preferable in terms of intercycle reproducibility. In any case, the diagnostic superiority of the rotational technique does not originate from phase-related overestimation of the degree of narrowing because the frames showing maximal severity of the stenoses were selected, for both the conventional and the rotational technique, irrespective of the phase of the cardiac cycle.
We used a cine rate of 50 frames/s to obtain maximal information from both the standard and the panoramic approach. Since a 180° rotation was accomplished in 4 s, the perspective gap between consecutive frames amounts to ∼1°. Although a question not specifically addressed in this study, theoretic and practical considerations indicate that a cine rate of 25 frames/s would have yielded similar results. In fact, 1) the corresponding perspective resolution of two degrees is much lower than the theoretic angle α lim, above which underestimation of the severity of the stenosis occurs; 2) the results of this study are based on the average of three consecutive frames and thus on perspective information spanning over 3°.
The geometry of the angiographic unit allows a maximal cranial and caudal tilt of ∼30° to be applied to a 180° transaxial scan, and this was actually limited to 25° in the present study. Although with the standard projections a greater amount of tilt (up to 45°) could be achieved, in no case did this yield greater diagnostic accuracy with respect to panoramic imaging.
In this study, quantitative angiography by a semiautomatic algorithm was used. A completely automatic, operator-independent technique such as the Coronary Artery Analysis System (CAAS) would have been preferable on theoretic grounds. However, the study involved comparison between frames obtained by two different approaches, necessarily unoptimized for quantitative analysis. On images randomly derived from routine arteriograms, fully automated arterial border detection was reported to be applicable to only 52% of the cases . Conversely, semiautomatic analysis could be successfully applied to all lesions and was selected as the only practicable alternative. Also, because quantitative analysis of routine arteriograms is in any case limited by operator-dependent variability in stenosis identification and frame selection , we elicited to select by consensus three consecutive frames showing maximal stenosis severity and used for comparisons the averaged estimates of all operators on all selected frames.
In this study, manual rotation of the Carm was accomplished because automated rotation at a comparable speed was not available. Obviously, a fully automated technique would be preferable. New systems equipped with a Carm rotating at 30°/s are now marketed by several manufacturers, and rapid, automated rotation allowing for conventional contrast injection will probably be available in a near future.
3.4 Safety Considerations
Although the overall dosage is reduced, the amount of contrast medium injected for each cine run is slightly increased with respect to the standard projections. This could conceivably result in an increased risk of contrast-induced complications, such as arrhythmias and prolonged ischemia. Despite this, no increase in complications could be detected during the experimental part of this study. Moreover, procedure log analysis of a subsequent series of >1,500 consecutive patients studied by the panoramic approach showed that the overall incidence of complications was below the accepted standards for coronary angiography .
The consequences of the new approach in terms of X-ray exposure were not directly measured in the present study. However, a shorter cine run conceivably implies a reduction of the dosage to both the patient and the personnel, provided that frame rate and exposure time are within the limits of the standard approach. This was applicable to all of the above series of 1,500 patients, where satisfactory films were consistently obtained using a frame rate of 25 frames/s and an exposure time <5.0 ms.
3.5 Alternative Approaches to Three-Dimensional Information
Although a three-dimensional reconstruction of the coronary arteries can be obtained by rotating or multiple-tube dedicated machines, such as the dynamic spatial reconstructor or flashing tomosynthesis computed tomography , these are costly and not widely available approaches. The panoramic technique can be accomplished by a standard angiographic unit, and the resulting images are recorded on a standard cine film, at no additional cost.
The diagnostic capability of this technique was intuitively recognized by Campeau and Saltiel , who proposed a rotational cineangiographic assessment of congenital heart disease >30 years ago.
Recently, neuroradiologic applications using essentially the same principle that we applied to the coronary arteries have been described. When compared with the standard digital subtraction angiogram, the rotational technique was reported to allow better visualization of vascular anatomy and to improve the angiographic assessment of cerebral aneurysms [30, 31].
This new technique can be safely implemented on a standard angiographic equipment. Use of contrast medium and cine film is reduced. The panoramic approach is operator independent and is expected to improve reproducibility among subsequent examinations from the same and different laboratories. In a series of 129 consecutive patients with coronary artery disease, we demonstrated improved diagnostic accuracy compared with the standard projections. This study suggests that two rotational scans with 25° cranial and caudal angulation could be used as an advantageous alternative to the traditional, multiview approach for routine diagnostic coronary angiography.
- left anterior oblique
- left lateral
- right anterior oblique
- Received December 3, 1996.
- Revision received October 23, 1997.
- Accepted December 18, 1997.
- The American College of Cardiology
- Grondin CM,
- Dyrda I,
- Pasternac A,
- Campeau L,
- Bourassa MG,
- Lesperance J
- Isner JM,
- Kishel J,
- Kent KM
- Vlodaver Z,
- Frech R,
- Van Tassel RA,
- Edwards JE
- Nissen SE,
- Gurley JC,
- Grines CL,
- et al.
- Mintz GS,
- Painter JA,
- Pichard AD,
- et al.
- Ramee SR,
- White CJ,
- Collins TJ,
- Mesa JE,
- Murgo JP
- Gurley JC,
- Nissen SE,
- Booth DC,
- DeMaria AN
- Thomas AC,
- Davies MJ,
- Dilly S,
- Dilly N,
- Franc F
- Gould KL,
- Kelly K,
- Bolson EL
- Kirkeeide RL,
- Gould KL,
- Parsel L
- Topol EJ,
- Nissen SE
- Bunnel IL,
- Greene DG,
- Tandon RN,
- Arani DT
- Reiber JHC,
- Serruys PW,
- Kooijman CJ,
- et al.
- Bittl JA,
- Levin DC
- Campeau L,
- Saltiel J
- Holff DJ,
- Wallace MC,
- terBrugge KG,
- Gentili F
- Tu RK,
- Cohen WA,
- Maravilla KR,
- et al.