Computed tomography coronary angiography (CTCA) is increasingly performed worldwide. For the interpretation of the acquired data set, different post-processing techniques are available, such as multiplanar reformation, maximum-intensity projections, direct volume rendering, virtual coronary angioscopy or the angiographic view. Each of these techniques shows certain advantages and disadvantages during application and image interpretation. Thus, a combination of post-processing techniques for the interpretation of CTCA studies should be used. When starting to perform and interpret CTCA, a systematic approach is mandatory for accurate diagnosis. We developed a practical algorithm in our institution for the interpretation of CTCA studies with special emphasis on interpretation steps to avoid a false-negative or false-positive diagnosis. In this article we discuss the strengths and weaknesses of the different post-processing techniques available for evaluation of CTCA and provide a systematic approach for interpreting a CTCA study, with an emphasis on how to avoid false-positive and false-negative classifications.
Computed tomography coronary angiography (CTCA) has entered the level of daily clinical practice in many institutions worldwide. All studies previously performed on the diagnostic accuracy of CTCA have shown a high negative predictive value, indicating a high ability of this method to exclude relevant coronary artery disease.1–3 However, CTCA is highly demanding not just of the technology, but also of the interpreters of the CTCA data sets. For the inexperienced reader a relevant coronary lesion could be easily missed or a non-relevant stenosis could be overestimated as a significant lesion, particularly in the presence of severe calcified deposits. In addition, artefacts might be mistaken for real lesions, resulting in avoidable false-positive classifications. It is of utmost importance that a CTCA study is correctly interpreted as the reported high negative predictive value of CTCA is one of the main strengths of this method, and a patient with a negative scan result will frequently not undergo further cardiac diagnostics. On the other hand, any false-positive results at CTCA result in further invasive work-up, which would have been avoided if the CTCA interpretation had been correct.
With improvements in spatial resolution, CTCA data sets have become increasingly large with about 1,000–5,000 images per examination. Therefore, simple transverse scanning of such data sets by interactively moving up and down a stack of axial slices is impractical and favours a shift towards volume imaging and 3D image display.4 Thus, the interpreter of a CTCA data set should be familiar with the advantages and disadvantages of the available post-processing techniques. Some of these post-processing techniques have a large number of possible parameters that may be tuned to obtain the best visualisation for a given data set. Interpretation of CTCA requires interactive manipulation and browsing of images. Therefore, the interpreter of CTCA can improve his or her reading by understanding the principles of the individual post-processing techniques and by learning to interact with his workstation.
Post-processing Techniques Used for Evaluation of Computed Tomography Coronary Angiography
The 3D course and the small size of the coronary arteries lead to unfavourable cutting angles and partial volume effects on the axial images. Different clinical tasks, such as the detection of atherosclerotic plaques, assessment of coronary artery stenosis or the evaluation of stents and coronary artery bypass, frequently require different visualisation techniques.
Finally, communication of findings to referring physicians is more efficient if the referrer is familiar with the given representation. For example, while cardiologists are used to conventional angiographic images, surgeons will prefer realistic 3D images simulating a surgical view. Therefore, stacks of axial images need to be integrated into a form that can be interpreted more efficiently and accurately and that appear similar to presentations already established in clinical practice.
Multiplanar reformation (MPR) is a 2D technique where only the data in a plane inside the original 3D image volume are displayed. This plane may be orthogonal to the axial slices of the image volume (i.e. coronal or sagittal MPR), oblique or replaced by an arbitrary curved surface (i.e. curved MPR) (see Figure 1). Oblique MPR can be used to show the four-chamber view and short-axis view. The luminal narrowing of identified plaques can be assessed by interactively reviewing plaques in planes longitudinal and orthogonal to the vessel centre-line. Curved MPR enables the 2D display of the complete course of a vessel in a single image by fitting a surface along the vessel centre-line. Instead of defining the vessel centre-line manually on axial slices, current dedicated workstations permit automatic defining of the centre-line after the user has selected a seeding point inside the lumen of the vessel. A drawback of the curved MPR technique is that only a single vessel branch can be displayed at a time (see Table 1).
MPR is the most important post-processing technique in CTCA, especially for the assessment of coronary stenosis in images reformatted perpendicular to the vessel centreline. Most importantly, among all post-processing techniques MPR is the least susceptible to incorrect operation by the radiologist. However, interactive viewing of the data sets and continuous adaptation of reformation planes is required since an improper plane orientation might introduce false-positive or false-negative vessel stenoses.5
A maximum-intensity projection (MIP) is the simplest 3D post-processing technique. For each pixel in the image a ray is cast through the 3D image volume and only the highest-attenuation voxels are retained, thereby allowing rapid display of part of or the total volume data set (see Figure 1). The only parameters are the window setting and the angulation of rays, which determinesthe viewing orientation.
MIP allows the display of coronary arteries including their side branches in a single image and is able to visualise smaller branches with less effort than is required compared with other 3D techniques.4 The significant information reduction by this technique also has some drawbacks (see Table 1). When the MIP technique is applied to the full image volume, usually the coronary arteries will overlap with the contrast-medium-filled cardiac cavities, high-attenuation pericardial metallic clips and/or the bones of the chest and therefore may be obscured by these structures. Thin-slab MIP is a simple technique to consider only a slab volume bounded by editing planes in front and behind a vessel,6 thereby removing the high-attenuation voxels of surrounding structures from the rendering (see Figure 1). Attention should be drawn to the need not to use too high a slab thickness of the MIP display (commonly less than 10mm) in order to avoid overestimation of coronary artery stenosis in the presence of extensively calcified deposits or coronary stents.5,7 Despite these shortcomings, MIP has become an important complementary tool for the visualisation of coronary arteries.
Direct Volume Rendering
Direct volume rendering (DVR) is motivated by a physically based model that defines the optical properties of voxel densities encountered along the ray. A so-called transfer function maps Hounsfield scalar values to colours and opacities that are then integrated along the viewing ray. By setting up a suitable transfer function, separate tissues can be classified and rendered from completely transparent to completely opaque. The image volume is usually shaded according to the illumination from external light sources to enhance 3D depth perception (see Figure 1).
DVR provides a good overview of the coronary artery anatomy and is able to produce realistic images suitable for reporting image findings to referring physicians, and colour images are popular to take home among patients. On the other hand, the multitude of parameter settings makes this technique time-consuming. Inappropriate settings may mask or artificially display a coronary lesion8 (see Table 1).
Virtual Coronary Angioscopy
Virtual coronary angioscopy (VCA) images are produced by DVR with the observer virtually positioned inside the vessel lumen (see Figure 1).
The definition of the vessel centre-line required for fly-through is automated in current software similar to automated centre-line definition in curved MPR. A high sensitivity for coronary plaque detection has been reported for VCA, and it may have some value in evaluating patency in the setting of heavily calcified vessels or stents.9,10 However, the clinical application of VCA has been questioned for several reasons: correct way-finding is highly dependent on superior image quality,11 it is difficult to orientate oneself in the VCA images, the post-processing is time-consuming and it suffers the same drawbacks as regular DVR, mainly that inappropriate settings may mask or artificially display a coronary lesion (see Table 1).
Angiographic view (AGV) images are an MIP of the CTCA data set after virtually removing the contrast-filled cardiac cavities and extra-cardiac structures (see Figure 1). AGV is able to display the entire coronary artery tree without surrounding tissue, displaying the CTCA comparable to images known from cardiac catheter, and thus is familiar to cardiologists. The drawbacks of AGV are that the post-processing is time-consuming and the technique shares the limitations of projectional luminography by overlapping the cardiac structures as known from cardiac catheter angiography (see Table 1).
Accuracy of the Different Post-processing Techniques
There are controversies about which post-processing technique is the most accurate for the identification of coronary plaques and the assessment of coronary stenosis. Previous studies have shown that transverse scanning is superior to the other post-processing techniques, mainly because it is least susceptible to cardiac motion artefacts.12 With the introduction of modern CT systems with increased spatial and temporal resolutions, recent studies have shown advantages of several post-processing techniques over transverse scanning.13,14 In our experience, MPR is the most relevant post-processing technique; however, optimal interpretation of CTCA studies should be based on the combination of two or more post-processing techniques.
Interpretation of Computed Tomography Coronary Angiography
CTCA data should be interpreted on a computer workstation at least capable of MPR, MIP, and DVR reconstructions. Figure 2 displays the standard algorithm for evaluation of CTCA. The algorithm has two interpretation steps: the first interpretation step for identification of coronary artery plaques is followed by a second interpretation step to avoid false-positive and false-negative classifications.
First Interpretation Steps
The first step in interpreting CTCA is the analysis of the axial source images. The reader is given a first overview about the image quality, the coronary artery anatomy and possible difficulties in interpretation to be expected by the presence of motion artefacts. By browsing through the axial images one should recognise all coronary plaques in the coronary artery tree. The main advantage of the axial source images is that these images are – in contrast to the post-processed images – not manipulated. Thus, the axial source images should be always used to confirm suspected pathologies found in the reformatted images. As an addition to the axial source images, thick-slab MIP (we prefer a slab thickness of 3–5mm) can be reformatted to illustrate the vessel of interest at a longer length.
Regions of plaques found at the inspection of the axial source images and the reformations need more detailed consideration on longitudinal and perpendicular projections to the vessel centreline to assess the degree of luminal diameter narrowing.
Second Interpretation Steps
After the first interpretation steps there are two possible results: if no relevant coronary artery stenosis is present, the CTCA data set should be reviewed again to avoid false-negative results; if a relevant coronary artery stenosis was found on the first interpretation, a second interpretation has to be performed in order to avoid false-positive results.
How to Avoid False-negative Findings
The algorithm for avoidance of false-negative classifications includes two steps. First, one should review areas with frequently high prevalence of relevant coronary artery stenosis. An evaluation of the cardiac CT studies of 800 patients with atypical chest pain and intermediate risk of coronary artery disease in our institution (unpublished data) revealed relevant coronary artery stenosis in 16% of patients (128 of 800).
Prevalence of relevant coronary stenosis is higher in the left anterior descending artery (LAD, 47%) than in the circumflex artery (CX, 28.5%) and the right coronary artery (RCA, 24.5%). On a segment-based analysis, the highest prevalence of coronary artery stenosis is observed in the proximal and mid-segments of the RCA, the proximal and mid-segments of the LAD and the proximal and distal segments of the CX. Two-thirds of coronary stenoses are found in these segments. Therefore, a special investigation of these areas with high prevalence of coronary artery stenosis is recommended.
The second step is to review areas where we know that we often tend to miss a lesion. There are three areas one should routinely review after the initial evaluation before concluding that the CTCA study is normal: the distal segment of the RCA and the origin of the posterior descending artery, the proximal and mid-segment of the LAD, including the origin of the first diagonal branch, and the distal segment of the CX near the origin of the obtuse marginal branch. We routinely review those areas by thick-slab MIP and DVR images to avoid missing a lesion. Note that the proximal LAD and the distal CX are also areas with a commonly high prevalence of relevant coronary artery stenosis.
How to Avoid False-positive Findings
One potential pitfall in the assessment of coronary artery stenosis is to mistake a motion artefact as a non-calcified plaque. This can particularly occur in CTCA data sets of reduced image quality. One should always check a second reconstruction time-point for the presence of this plaque. If the plaque is only seen on one of the reconstruction time-points, a motion artefact has to be expected simulating a pathology (see Figure 3).
Another important subject for distinguishing motion artefacts from real lesions is that any plaque has to be visible on the reconstructed images, i.e. that the plaque could be differentiated from the surrounding pericardial tissue and the contrast in the lumen (see Figure 4). Otherwise, the ‘lesion’ has to be suspected as artefact.
For all identified and verified plaques, the relevance of luminal narrowing has to be evaluated. The degree of coronary stenosis is calculated as a ratio of the luminal diameter at the site of stenosis compared with a normal-appearing reference site in an adjacent proximal or distal vessel segment. The most common cause of false-positive classifications at CTCA is the presence of coronary artery calcifications. Such high-density structures cause beam-hardening and blooming artefacts resulting in an overestimation of the degree of stenosis. In the presence of calcifications we routinely reconstruct an additional data set using a sharp convolution kernel and prefer the evaluation of this data set using a bone window setting (window width 1400HU, window centre 500HU; see Figure 5).
Reporting of Imaging Findings
We use a standardised template for reporting CTCA findings and prefer the inclusion of schematic drawings as a synopsis of the relevant findings and representative images of examination. The first part in the finding section is dedicated to the results of the calcium scoring. The distribution is to be characterised as located or diffuse, and the extent as nodular or massive.
The coronary anatomy, the presence of anomalies and the coronary artery supply type (right-dominant, left-dominant or balanced supply type) should be mentioned in the report. Each coronary artery (i.e. left main artery, RCA, LAD and CX) should be separately commented on with regard to the presence and type of coronary artery stenosis. We use the term mild stenosis for a luminal diameter narrowing of less than 30%, moderate for 30–50%, moderate to severe for 50–75% and severe stenosis for more than 75%. If present, coronary artery bypass grafts are described with type of graft, origin, course, site of anastomosis and the native coronary arteries. Location and degree of graft stenosis have to be indicated in the report.
The location of stenosis is reported using the terminology of catheter angiography reports that describe the position as in the proximal, mid- or distal segment of the affected vessel or their side branches. The description of cardiac structures includes the ventricular function analysis, the general cardiac anatomy, anomalies of the myocardium and variations. In a separate section extra-cardiac findings of the mediastinal structures, the lungs and the chest wall are described.
We provide image samples to the referring physician, including curved reconstructions along the main coronary arteries and images displaying all relevant findings.
Cardiac CT offers many advantages to patients and referrers by providing an accurate and comprehensive evaluation not only of the coronary arteries but also of the surrounding cardiac and extra-cardiac structures. Variations in cardiac and coronary anatomy and the increasing number of examinations performed worldwide require a standardised systematic approach for interpretation of cardiac CT studies, the awareness of potential pitfalls to minimise the rate of false-negative and false-positive interpretations and knowledge of the strengths and weaknesses of the different post-processing techniques available for evaluation of CTCA.
- Leschka S, Stolzmann P, Desbiolles L, et al., Diagnostic accuracy of high-pitch dual-source CT for the assessment of coronary stenoses: first experience, Eur Radiol, 2009;19:2896–2903.
Crossref | PubMed
- Leschka S, Stinn B, Schmid F, et al., Dual source CT coronary angiography in severely obese patients: trading off temporal resolution and image noise, Invest Radiol, 2009;44:720–7.
Crossref | PubMed
- Maffei E, Palumbo A, Martini C, et al., Diagnostic accuracy of 64-slice computed tomography coronary angiography in a large population of patients without revascularisation: registry data and review of multicentre trials, Radiol Med, 2010;115(3):368–84.
Crossref | PubMed
- Fishman EK, Ney DR, Heath DG, et al., Volume rendering versus maximum intensity projection in CT angiography: what works best, when, and why, Radiographics, 2006;26: 905–22.
Crossref | PubMed
- van Ooijen PMA, Ho KY, Dorgelo J, Oudkerk M, Coronary artery imaging with multidetector CT: visualization issues, Radiographics, 2003;23:e16.
Crossref | PubMed
- Napel S, Rubin GD, Jeffrey RB Jr., STS-MIP: a new reconstruction technique for CT of the chest, J Comput Assist Tomogr, 1993;17:832–8.
Crossref | PubMed
- Anderson CM, Saloner D, Tsuruda JS, et al., Artifacts in maximum-intensity-projection display of MR angiograms, AJR Am J Roentgenol, 1990;154:623–9.
Crossref | PubMed
- Lu B, Dai RP, Jiang SL, et al., Effects of window and threshold levels on the accuracy of three-dimensional rendering techniques in coronary artery electron-beam CT angiography, Acad Radiol, 2001;8:754–61.
Crossref | PubMed
- Schroeder S, Kopp AF, Baumbach A, et al., Noninvasive detection and evaluation of atherosclerotic coronary plaques with multislice computed tomography, J Am Coll Cardiol, 2001;37:1430–5.
Crossref | PubMed
- Lawler LP, Pannu HK, Fishman EK, MDCT evaluation of the coronary arteries, 2004: how we do it—data acquisition, postprocessing, display, and interpretation, AJR Am J Roentgenol, 2005;184:1402–12.
Crossref | PubMed
- van Ooijen PM, de Jonge G, Oudkerk M, Coronary fly-through or virtual angioscopy using dual-source MDCT data, Eur Radiol, 2007;17:2852–9.
Crossref | PubMed
- Vogl TJ, Abolmaali ND, Diebold T, et al., Techniques for the detection of coronary atherosclerosis: multi-detector row CT coronary angiography, Radiology, 2002;223: 212–20.
Crossref | PubMed
- Ferencik M, Ropers D, Abbara S, et al., Diagnostic accuracy of image postprocessing methods for the detection of coronary artery stenoses by using multidetector CT, Radiology, 2007;243:696–702.
Crossref | PubMed
- Jinzaki M, Sato K, Tanami Y, et al., Diagnostic accuracy of angiographic view image for the detection of coronary artery stenoses by 64-detector row CT: a pilot study comparison with conventional post-processing methods and axial images alone, Circ J, 2009;73:691–8.