Author + information
- Received October 1, 2007
- Revision received December 19, 2007
- Accepted January 6, 2008
- Published online February 3, 2009.
- Gudrun M. Feuchtner, MD, PD⁎,⁎ (, )
- Paul Stolzmann, MD§,
- Wolfgang Dichtl, MD, PhD, PD†,
- Thomas Schertler, MD§,
- Johannes Bonatti, MD, FECTS‡,
- Hans Scheffel, MD§,
- Silvana Mueller, MD†,
- André Plass, MD∥,
- Ludwig Mueller, MD‡,
- Thomas Bartel, MD, PD†,
- Florian Wolf, MD¶ and
- Hatem Alkadhi, MD, PD§
- ↵⁎Reprint requests and correspondence:
Dr. Gudrun M. Feuchtner, Innsbruck Medical University, Department of Radiology II, Anichstrasse 35, A-6020 Innsbruck, Austria
Objectives The aim of this study was to assess the value of multislice computed tomography (CT) for the assessment of valvular abnormalities in patients with infective endocarditis (IE) in comparison with transesophageal echocardiography (TEE) and intraoperative findings.
Background Multislice CT has recently shown promising data regarding valvular imaging in a 4-dimensional fashion.
Methods Thirty-seven consecutive patients with clinically suspected IE were examined with TEE and 64-slice CT or dual-source CT. Twenty-nine patients had definite IE and underwent surgery.
Results The diagnostic performance of CT for the detection of evident valvular abnormalities for IE compared with TEE was: sensitivity 97%, specificity 88%, positive predictive value (PPV) 97%, and negative predictive value (NPV) 88% on a per-patient basis (n = 37; excellent intermodality agreement κ = 0.84). CT correctly identified 26 of 27 (96%) patients with valvular vegetations and 9 of 9 (100%) patients with abscesses/pseudoaneurysms compared with the intraoperative specimen. On a per-valve–based analysis, diagnostic accuracy for the detection of vegetations and abscesses/pseudoaneurysms compared with surgery was: sensitivity 96%, specificity 97%, PPV 96%, NPV 97%, and sensitivity 100%, specificity 100%, PPV 100%, NPV 100%, respectively, without significant differences as compared with TEE. Vegetation size measurements by CT correlated (r = 0.95; p <0.001) with TEE (mean 7.6 ± 5.6 mm). The mobility of vegetations was accurately diagnosed in 21 of 22 (96%) patients with CT, but all of 4 leaflet perforations (≤2 mm) were missed. CT provided more accurate anatomic information regarding perivalvular extent of abscess/pseudoaneurysms than TEE.
Conclusions Multislice CT shows good results in detecting valvular abnormalities in IE and could be applied in pre-operative planning and exclusion of coronary artery disease before surgery.
Infective endocarditis (IE) represents a severe disease with a high mortality rate of up to 40% (1). Despite continuous advances in medical health care, the prevalence of IE has not declined over the past 30 years. In contrast, a recent rise in the prevalence of IE has been reported and explained by increasing predisposing factors such as degenerative valvular disease, implanted pacemakers and defibrillators, and prosthetic valves, all of which are enhanced by a general increase in the elderly population.
In clinical practice, the diagnosis of IE is often difficult, and both overdiagnosis and underdiagnosis are observed (2). The diagnosis of IE is established using the modified Duke criteria based on clinical and imaging findings (3). Echocardiography represents the central tenant in evaluating patients who have a clinical presentation that raises the concern for IE. With echocardiography, there are several findings that provide evidence of IE, including vegetations, evidence of periannular tissue destruction, abscesses, aneurysms, fistulas, leaflet perforations, or valvular dehiscence. Transthoracic echocardiography (TTE) has shown limited sensitivity for the diagnosis of IE ranging between 18% and 63% (4) with a detection rate of 25% for small vegetations <5 mm (5). Therefore, transesophageal echocardiography (TEE) is usually also performed. Based on its higher specificity and sensitivity ranging between 48% to 100% (4), TEE is considered the reference imaging method. Nevertheless, both TTE and TEE are limited by their dependence on the individual patient's morphology, instrumental settings, transducer position, operator, and artifacts from heavy valve calcifications and metallic prosthetic valves through acoustic shadowing. Therefore, if an initial echocardiography is negative, repeated TEE examinations are recommended in cases of clinical suspicion (6,7).
Recent advances in the temporal and spatial resolution of multislice computed tomography (CT) scanners allow high-resolution cardiac imaging. Currently, the major application of multislice CT is the evaluation of coronary artery disease to exclude coronary stenosis >50% (8,9). In addition, CT has recently shown promising results in valvular imaging (10–13) by providing morphologic information and functional 4-dimensional cine-imaging.
Thus, cardiac multislice CT could qualify as an alternative imaging modality for visualizing valvular abnormalities in IE, and therefore, the purpose of this study was to compare CT with TEE and intraoperative findings.
Between August 2006 and September 2007, 37 consecutive patients with suspected IE were examined. The diagnosis of IE was based on the modified Duke criteria (2,3). According to Haldar and O'Gara (2), “definitive IE” (n = 29) was defined as the presence of either 2 major criteria or 1 major and 3 minor criteria. “Possible IE” (n = 8) was defined as the presence of either 1 major (positive hemoculture with specific bacteria) and 1 minor (n = 4) or 3 minor criteria (n = 4) out of the following: positive hemoculture with nonspecific microorganism, fever >38°C, predisposing cardiac conditions (e.g., prosthesis, valvular disease) or intravenous drug abuse, immunologic phenomena, or vascular phenomena.
CT was performed on the same day as (n = 27) or 1 day after (n = 10) echocardiography. Twenty-eight patients underwent surgery within 5 days after CT, 1 patient was operated on 6 weeks later because of comorbidities.
All cardiac CT examinations were clinically indicated. Reasons for referrals to CT were exclusion of coronary artery disease (n = 31) and/or uncertain echocardiography findings regarding anatomic extent of extravalvular involvement (n = 12) and the assessment of coronary bypass patency (n = 1). Patient exclusion criteria were renal dysfunction (serum creatinine >1.5 mg/dl), hyperthyroidism, iodine allergy, New York Heart Association functional class III to IV heart failure, and pregnancy. Within the time period of data acquisition, 8 patients were excluded because of nephropathy (n = 4), iodine allergy (n = 2), New York Heart Association functional class IV heart failure (n = 1), and emergency surgery before the CT scan (n = 1). The institutional review board approved the study, and written informed consent was obtained.
CT Scan Protocol
Twenty patients underwent imaging with a 64-slice CT scanner (Sensation 64, Siemens, Munich, Germany) using the following parameters: detector collimation 32 × 0.6 mm, slice acquisition 64 × 0.6 mm by means of z-flying focal spot, pitch 0.2, gantry rotation time 330 ms, tube voltage 120 kV, and tube-current-time product 600 to 800 mAs.
Seventeen patients underwent imaging with a dual-source CT scanner (Somatom Definition, Siemens): detector collimation 2 × 32 × 0.6 mm, slice acquisition 2 × 64 × 0.6 mm, pitch 0.2 to 0.53 (depending on heart rate), gantry rotation time 330 ms, tube voltage 120 kV, and tube-current-time product 350 mAs/rotation.
With 64-slice CT, a beta-blocker was administered intravenously (5 to 10 mg metoprolol) before the examination in patients with heart rates >75 beats/min; no heart rate control was performed in the patients undergoing dual-source CT.
The scan direction was cranio-caudal during mid-inspiration. The scan ranged from pulmonary artery bifurcation level to the heart base. The echocardiogram signal was recorded simultaneously, and echocardiogram-gated dose modulation was used.
A bolus of 80 to 100 ml of iodinated contrast agent (Iodixanol, Visipaque 320, GE Healthcare, Buckinghamshire, United Kingdom) followed by a 40-ml saline flush was intravenously injected into an antecubital vein at a flow rate of 5 ml/s. The scan delay was calculated using the bolus-tracking technique (ascending aorta, threshold 100 HU). A dedicated contrast agent injection protocol for both left and right ventricular enhancement was applied in patients with pacemakers (14).
CT Image Reconstruction
Two transverse image sets (A and B) were reconstructed (slice thickness 0.75 mm, increment 0.4 mm, medium-smooth convolution kernel, retrospective electrocardiogram gating during mid-systole to late systole [10% to 40%] and mid-diastole to late diastole [60% to 80%]). Another dataset (C) containing transverse images was reconstructed in 10% steps from 0% to 90% of the cardiac cycle (slice thickness 1.0 mm, increment 0.7 mm).
CT Image Post-Processing and Analysis
The post-processing was performed on a separate workstation (Leonardo, Siemens).
The valves were evaluated according to the following echocardiographic criteria (4):
1. Vegetations, defined as irregularly shaped, oscillating masses, adherent to and distinct from the endocardium; their sizes were measured on multiplanar reformations using an electronic caliper.
2. Abscesses, defined as irregularly shaped, inhomogeneous paravalvular masses within periannular region, myocardium, or pericardium; a pseudoaneurysm was defined as a space filled with contrast medium with a communication with the cardiac chambers or the aortic root.
3. Fistula, defined as a continuation between the chambers of the left and right heart.
4. Leaflet perforation, defined as a discontinuation.
5. Valvular dehiscence, defined as a rocking motion of a prosthetic valve with an excursion >15° in any 1 plane.
Image sets A and B were analyzed on multiplanar reformations and by volume rendering technique. Image set C was used for 4-cine dynamic imaging. The CT data analysis was performed by 2 experienced, independent readers blinded to the echocardiography. For intraobserver agreement of CT findings, 2 readers reanalyzed the blinded data after 2 months, and consensus readout was appended in the case of disagreement.
Coronary arteries were evaluated with regard to significant coronary stenosis >50%. The valves were assessed regarding aortic stenosis (10) or regurgitation (11), and mitral stenosis (12) or regurgitation (13).
The TEE examinations were performed by 2 experienced reviewers according to a standardized database/reporting system. A multiplane 5-MHz TEE probe (Sonos 5500, Philips Company, Andover, Massachusetts) was used in 20 patients, a multiplane 2- to 7-MHz TEE transducer (Acuson Sequoia 512, Acuson-Siemens Medical Systems, Munich, Germany) in 17 patients. Doppler B- and M-mode capabilities were applied. Echocardiographers were blinded to the results from CT but fully aware of the clinical history.
Twenty-nine patients underwent cardiac surgery. The indication of surgery was based on the American Heart Association guidelines (15). The surgeons evaluated valve abnormalities by direct visualization, digital inspection, and instrumental exploration. Surgical criteria for IE were vegetations, leaflet perforations, and/or abscesses, which were subsequently confirmed by histopathology. Perivalvular involvement was evaluated regarding annulus destruction and/or myocardial and pericardial involvement. Surgery consisted of valve replacement or repair in patients with infection limited to the valvular leaflets or of radical resection of infected paravalvular tissues and reconstruction with patches and valve replacement in patients with abscesses.
The first part of the statistics assessed the value of CT for the diagnosis of IE in comparison with TEE. All patients with definite and possible IE (2,15) were included. Surgery as validation of the imaging findings was not available because patients with possible IE but negative imaging findings were not operated on. The second part included assessment of the diagnostic performance of CT and TEE in comparison with the gold standard, surgery.
The diagnostic accuracy of CT was calculated, and differences among CT, TEE, and intraoperative findings were tested for significance by using the McNemar test. Intraobserver agreement for CT readout and intermodality agreement were tested with Cohen's kappa statistics (16). Pearson correlation coefficients were used to compare CT measurements of vegetations with TEE. A value of p < 0.05 was considered significant for all tests. The statistical analysis was performed using SSPS version 12.0 (SSPS Inc., Chicago, Illinois).
Multislice CT was successfully performed in all 37 patients. Patient demographics and clinical data are listed in Table 1.
Infective endocarditis involved the mitral and/or aortic valve in 28 of 29 (97%) patients with definite IE. In 1 (3%) patient, IE involved the tricuspid valve and the right ventricular pacemaker lead. Thus, a total of 57 valves (28 mitral, 28 aortic, and 1 tricuspid) were included in the comparative analysis between CT and TEE, taking intraoperative findings as gold standard. Fifty-one of 57 (89%) valves were native, and 6 of 57 (11%) were prosthetics (3 of 6 mechanical and 3 of 6 bioprosthetic).
The results of intraoperative findings compared with CT are shown in Table 2.
Intraobserver and interobserver readout of CT findings
The interobserver agreement for diagnosis of IE (vegetation and/or abscess) per patient was κ = 0.94, and the intraobserver agreement was κ = 1. For the detection of vegetations, abscesses, fistulae, and mobility of vegetations, the intraobserver agreement was κ = 1 and the interobserver agreements were κ = 0.96, κ = 1, κ = 1, and κ = 0.94, respectively. Because of the low interobserver variability, the consensus reading was taken for all subsequent analyses.
CT versus TEE in patients with possible and definite IE
In 37 patients with possible or definite IE, CT identified 28 of 29 (97%) patients who had valve abnormalities that were identified by TEE (agreement: κ = 0.84).
CT correctly classified 70 of 73 (96%) valves to have abnormalities indicating IE or not (intermodality agreement: κ = 0.92). The diagnostic accuracy of CT compared with TEE on the per-patient– and per-valve–based analysis is shown in Table 3.
CT versus TEE for size and mobility of vegetations and dehiscence
Mean size of vegetations at CT (7.1 ± 5.8 mm, range 2 to 22 mm) significantly correlated (r = 0.95, p < 0.001) with TEE (7.6 ± 5.6 mm, range 1 to 22 mm).
The mobility of valvular vegetations was diagnosed by CT in 22 of 23 (96%) patients by applying 4-dimensional cine imaging (intermodality agreement: κ = 0.96).
CT correctly diagnosed 1 dehiscent mechanic aortic valve prosthesis.
CT and TEE versus surgery for vegetations
In 29 patients, CT correctly identified 26 of 27 (96%) patients who had intraoperatively proven vegetations (agreement: κ = 0.78).
CT correctly classified 55 of 57 (96%) valves to be either affected by vegetations or not, as confirmed during surgery (agreement: κ = 0.93). On a per-valve–based analysis, the diagnostic accuracy of CT compared with surgery was: sensitivity 96%, specificity 97%, PPV 96%, NPV 97% (Table 3).
CT depicted 41 of 46 (89%) vegetations when compared with intraoperative findings, but 5 vegetations were missed. One of these 5 vegetations (3 mm) was missed by CT because of artifacts. CT missed 3 small mitral valve vegetations (≤4 mm). One vegetation of the tricuspid valve was missed by CT because of low right ventricular contrast enhancement and metallic artifacts of a right ventricular pacemaker lead. One mobile aortic valve vegetation suspected by TEE was a mobile calcification by CT confirmed during surgery. Two vegetations on a mechanical prosthesis were correctly diagnosed with CT but were not depicted by TEE because of metal artifacts. One degenerative mitral valve leaflet thickening in a patient with aortic valve endocarditis was falsely classified as additional vegetation by CT. One aortic valve prolapse of a right coronary cusp was mistaken as a vegetation.
TEE correctly identified 26 of 27 (96%) patients with intraoperatively proven vegetations. The diagnostic accuracy of TEE for diagnosis of vegetations on a per-patient– (κ = 0.78) and per-valve– (κ = 0.97) based analysis compared with surgery was: sensitivity 96%, specificity 100%, PPV 100%, NPV 67%, and 96%, 100%, 100%, 97%, respectively.
Slight differences in accuracy for the detection of vegetations were found between CT and TEE on a per-valve–based analysis without a statistically significant difference (p = NS), and accuracy of vegetation detection on a per-patient basis was similar.
CT and TEE versus surgery for abscess and pseudoaneurysm in patients with definite IE
Nine of 9 (100%) paravalvular abscesses and/or pseudoaneurysms were correctly identified by CT compared with surgery (7 of 9 combined abscesses/pseudoaneurysms, 1 isolated aortic root abscess, and 1 aortic root pseudoaneurysm).
On a per-valve–based analysis, the diagnostic performance of CT for the detection of abscesses/pseudoaneurysms combined was: sensitivity 100%, specificity 100%, PPV 100%, and NPV 100%.
TEE correctly revealed 8 of 9 (89%) paravalvular abscesses and pseudoaneurysms, whereas 1 pseudoaneurysm was missed with TEE. The diagnostic accuracy of TEE for the detection of paravalvular abscesses and pseudoaneurysms in comparison with the intraoperative findings on a per-valve–based analysis was: sensitivity 89% (95% confidence interval [CI]: 52% to 100%), specificity 100% (95% CI: 93% to 100%), PPV 100% (95% CI: 63% to 100%), and NPV 98% (95% CI: 89% to 100%).
Slight differences in accuracy of CT and TEE for abscess and pseudoaneurysm detection were found on a per-valve–based analysis without statistical significance (p = NS).
CT was superior to TEE for showing the perivalvular involvement of abscesses and pseudoaneurysms in 3 patients. CT more often correctly showed the myocardial (n = 6 by CT vs. n = 3 by TEE), pericardial (n = 1 by CT vs. n = 0 by TEE), and coronary sinus involvement (n = 1 by CT vs. n = 0 by TEE) compared with echocardiography. CT correctly identified annulus involvement in 6 of 9 (67%) patients and correctly ruled out annulus involvement in 1 patient in whom extravalvular involvement was inconclusive at TEE. In 3 patients, pseudoaneurysms were located close to the right coronary artery (RCA). In 1 patient, the RCA could not be cannulated with invasive angiography, and exclusion of coronary artery disease was based on CT alone. In the other 2 patients, 3-dimensional visualization of the extent of pseudoaneurysms was useful for pre-operative planning of RCA bypass grafting.
CT and TEE versus surgery for fistula and cusp perforation
One of 26 (4%) patients had a fistula from the left ventricular outflow tract through the membranous ventricular septum to the right ventricle correctly diagnosed with both CT and TEE.
None of the 4 (0%) intraoperatively confirmed mitral valve leaflet perforations (<2 mm) could be visualized with CT but were detected by TEE.
Table 4 shows the prevalence of coronary artery disease and valvular disease, including quantitative measurements by CT.
Pre-operative assessment of coronary artery disease with conventional coronary angiography has been recommended in the guidelines of the American College of Cardiology/American Heart Association for selected patients scheduled for cardiac valve surgery (15). Recently introduced multislice CT allows noninvasive coronary artery imaging and has been shown to reliably exclude significant coronary artery stenosis (8,9). The major advantage of multislice CT over conventional coronary angiography in patients with acute IE is its noninvasiveness and thus avoidance of the high embolization risk of valvular vegetations during catheter manipulation. Once spread into the systemic arterial circulation, those vegetations can cause severe complications such as cerebral or peripheral embolization (17).
In this study, multislice CT showed an excellent diagnostic performance in assessing morphologic abnormalities commonly seen in patients with IE. Accuracy of vegetation, abscess, and pseudoaneurysm detection with CT was similar to that of the clinical reference standard imaging modality, TEE. Size measurements of vegetations with CT showed a good concordance with TEE, and 4-dimensional cine-mode allowed assessment of the mobility of vegetations, both of which represent risk factors for systemic arterial emboli (17–19).
CT missed some small vegetations (≤4 mm), which may be attributed to the limited temporal resolution of multislice CT. Moreover, false positive findings occurred with CT (e.g., by misclassifying a degenerative lesion). However, CT correctly detected vegetations attached to a mechanic prosthetic valve that were missed by TEE. A similar experience has been recently reported (20) for a pannus causing mechanic prosthetic valve dysfunction. Moreover, multislice CT was helpful in distinguishing between calcified degenerative lesions and vegetations.
All 4 mitral valve leaflet perforations (≤2 mm) were missed with CT, most probably related to lack of flow information that facilitates visualization of perforations with TEE.
CT correctly detected paravalvular abscesses and pseudoaneurysms in all patients, whereas TEE missed 1. Perivalvular abscesses develop in up to 30% of patients with IE (21), and visualization with TTE is often difficult, leading to delayed diagnosis (22). The early detection of perivalvular abscesses is crucial because the mortality of patients increases up to 2-fold (22). Sensitivity for the detection of abscesses associated with IE has been reported to be only 28% for TTE and 87% for TEE (22). In our study, multislice CT provided better anatomic information than TEE with regards to the perivalvular extent such as myocardial, pericardial, and coronary sinus involvement, which was helpful for planning surgery (e.g., to avoid unsuccessful attempts of retrograde cannulation of the coronary sinus). Additionally, 3-dimensional visualization of periannular tissue destruction and the extent and anatomic relationships of a perivalvular pseudoaneurysm was helpful for pre-surgical planning (e.g., to anticipate the anastomotic site of a bypass graft, avoid injury of the circumflex artery in cases of paravalvular mitral valve abscess during abscess debridement, and predict the risk of a ventriculoatrial disconnection [“hole-in-one”]).
The use of iodinated contrast media is contraindicated in patients with pre-existing renal dysfunction. Another limitation of multislice CT is the radiation exposure ranging from 9 to 11 mSv for 64-slice CT (23) and 7 to 9 mSv (24) for dual-source CT when echocardiogram pulsing for radiation dose reduction is used. However, cardiac CT allows comprehensive noninvasive assessment of coronary artery disease, and invasive coronary angiography can be avoided. Given the high mortality and morbidity of patients with acute IE, radiation exposure may be less concerning.
Absolute arrhythmia such as atrial fibrillation represents a limitation of echocardiogram-gated multislice CT, leading to misalignment artifacts. However, high-temporal resolution dual-source CT may be considered in patients with normofrequent atrial fibrillation.
Cardiac CT usually is aimed at visualizing the coronary arteries, including the left atrium, left ventricle, and mitral and aortic valves. Hence, valvular abnormalities associated with IE involving the pulmonary and tricuspid valves require an adaptation of the contrast media protocol (14) to achieve both left and right ventricular enhancement.
Furthermore, the statistical power of the “per-patient–based analysis” to exclude IE is limited because of low patient number. However, the per-valve–based analysis included a high number of “negative” valves. Further prospective studies on consecutive patients with “possible IE” would be desirable to determine the accuracy of CT to rule out IE in patients with clinical suspicion but negative TEE.
Results from this preliminary study show a good diagnostic value of multislice CT for the detection of valvular abnormalities associated with IE. Vegetations, abscesses, and pseudoaneurysms were detected with a similar accuracy with CT compared with TEE. CT could be used as an alternative imaging tool in patients with clinical suspicion of IE after an initial negative or inconclusive TEE. In detail, these could be patients in whom a chest CT scan is clinically indicated because of fever of unknown origin in the presence of other pre-disposing factors for IE (3 minor criteria) or a positive hemoculture (1 major).
CT might be also helpful if metallic artifacts hamper the visualization of prosthetic valves by TEE.
CT can be applied in clinical practice for the pre-operative ruling out of coronary artery disease instead of invasive angiography, if indicated (e.g., because of coexistent aortic stenosis or clinically suspected coronary artery disease). Moreover, comprehensive evaluation of anatomic relationships of perivalvular abscess/pseudoaneurysms may be beneficial for pre-surgical planning.
For accompanying videos and legends, please see the online version of this article.
This research was supported by the National Center of Competence in Research, Computer Aided and Image Guided Medical Interventions of the Swiss National Science Foundation.
- Abbreviations and Acronyms
- confidence intervals
- computed tomography
- infective endocarditis
- negative predictive value
- positive predictive value
- right coronary artery
- transesophageal echocardiography
- transthoracic echocardiography
- Received October 1, 2007.
- Revision received December 19, 2007.
- Accepted January 6, 2008.
- American College of Cardiology Foundation
- Erbel R.,
- Rohmann S.,
- Drexler M.,
- et al.
- Sochowski R.A.,
- Chan K.L.
- Vieira M.L.C.,
- Grinberg M.,
- Pomerantzeff P.M.A.,
- Andrade J.A.,
- Mansur A.J.
- Meijboom W.B.,
- Mollet N.R.,
- Van Mieghem C.A.,
- et al.
- Feuchtner G.M.,
- Dichtl W.,
- Friedrich G.J.,
- et al.
- Messika-Zeitoun D.,
- Serfaty J.M.,
- Laissy J.P.,
- et al.
- Bonow R.O.,
- Carabello B.A.,
- Chatterjee K.,
- et al.
- Vilacosta I.,
- Graupner C.,
- San Román J.A.,
- et al.
- Di Salvo G.,
- Habib G.,
- Pergola V.,
- et al.
- Leborgne L.,
- Renard C.,
- Tribouilloy C.
- Rohmann S.,
- Erbel R.,
- Mohr-Kahaly S.,
- Meyer J.
- Hausleiter J.,
- Meyer T.,
- Hadamitzky M.,
- et al.