The detection of a thrombus at the left atrial appendage (LAA) is an important step for management in a patient with a suspected embolic infarction. However, spontaneous echocardiographic contrast (SEC), which can mimic thrombus, can confuse clinicians in many cases. We examined electrocardiographic-gated 64-slice multidetector computed tomography with a 2-phase scan and transesophageal echocardiography in 314 patients with suspected embolic stroke. The transesophageal echocardiographic findings were classified using a 5-grade scale and the multidetector computed tomographic findings were categorized as no filling defect, an early filling defect (a filling defect seen on early-phase images without considering the late-phase images), and a persistent filling defect (a filling defect seen on added late-phase images, as well as on early-phase images). For quantitative analysis, the ratio of Hounsfield units in the LAA to the ascending aorta (AA) was calculated for each early-phase and late-phase image (LAA/AA L ). Using transesophageal echocardiography as the reference standard, for no filling defect seen on early-phase images, the presence of a thrombus, including severe SEC, could be ruled out with 100% sensitivity and a 100% negative predictive value. When considering the addition of late-phase images, all persistent filling defects had resulted from the presence of a thrombus and severe SEC. However, using the optimal cutoff value of 0.5 for the LAA/AA L ratio, thrombi could be distinguished from severe SEC where all thrombi had a LAA/AA L ratio <0.5. In conclusion, our findings suggest that 2-phase multidector computed tomography is useful for the detection and differentiation of a thrombus from SEC at the LAA in patients with suspected embolic stroke.
Multidetector computed tomography (MDCT) has been regarded as an emerging noninvasive imaging modality for the detection of left atrial appendage (LAA) thrombi. However, previous studies have suggested that early-phase imaging had limited potential to distinguish a thrombus from spontaneous echocardiographic contrast (SEC) and that late-phase imaging might be able to delineate the 2 phenomena. To date, a paucity of data is available regarding the diagnostic performance of 2-phase MDCT for the differentiation of SEC and a thrombus at the LAA. The purpose of the present study was to evaluate the significance of 2-phase MDCT for the detection and differentiation of a thrombus from SEC in patients with suspected embolic stroke.
Methods
From December 2007 to May 2009, we prospectively and consecutively enrolled 314 patients who were admitted to our hospital with a diagnosis of acute stroke within 7 days of onset. The patients were required to meet any of the following inclusion criteria: (1) any cardiac evidence of embolic stroke, including known high-risk cardiac sources such as atrial fibrillation, sick sinus syndrome, valve disease, including mechanical heart valves, symptomatic congestive heart failure with <30% left ventricle ejection fraction and dilated cardiomyopathy ; (2) radiologic evidence of embolic stroke such as multiple acute infarctions closely related in time within both right and left anterior circulations or both anterior and posterior circulations, multiple scattered acute infarctions without evidence of significant steno-occlusion of the relevant arteries, or a single infarction at the corticosubcortical, purely cortical, or subcortical area >15 mm in diameter and without evidence of steno-occlusion in the relevant vessels ; and (3) clinical evidence of embolic stroke such as history of ischemic stroke or transient ischemic attack involving vascular territory different from the index stroke without significant steno-occlusion of any cerebral vessels or evidence of systemic embolism. All patients underwent routine transthoracic echocardiography first. Subsequently, transesophageal echocardiography (TEE) and/or cardiac MDCT were attempted within a 1-week interval. The patients who did not undergo TEE or cardiac MDCT were excluded from the present study. Our institutional review board approved the study, and all participating patients provided written informed consent.
Patients with a heart rate >70 beats/min received 10 to 30 mg intravenous esmolol (Jeilopharm, Seoul, Korea) before MDCT. All patients were scanned using a 64-slice MDCT scanner (Brilliance 64, Philips Medical Systems, Best, The Netherlands). A standard scanning protocol for early-phase imaging was used as previously described. Late-phase imaging using prospective electrocardiographic gating for added images of the LAA was also obtained 2 minutes after starting the injection of contrast media. To reduce the radiation dose, we used 64 × 0.625 mm, 120 kV, and 200 mA to encompass the ascending aorta to the middle of the left ventricle for the added images of the LAA. The calculated radiation dose was 7 to 11 mSv for early-phase imaging and 1.5 to 2.8 mSv for late-phase imaging, depending on the scanning range and the patient’s body weight.
Image data sets were reviewed at a commercially available workstation (Brilliance, Philips Medical Systems) with curved multiplanar reconstruction images with a 2-mm section thickness without a gap along the curved line of the long axis of the LAA, as well as axial and multiplanar images.
A 7-MHz multiplane transesophageal echocardiographic probe was used in conjunction with an echocardiographic imaging system (Acuson Sequoia 512, Siemens, Erlangen, Germany). The examination was performed according to the use of a well-standardized protocol. All images were recorded as movie images on digital videotape in real-time for display and evaluation. Images of the left atrium and LAA were evaluated in the horizontal (0°) plane and the most available plane obtained by rotating the imaging sector from 0° to 180° during continuous visualization of the LAA.
For qualitative analysis, 2 experienced radiologists (EJC and SIC) reviewed all the images from MDCT independently. Both radiologists were unaware of the clinical information, and any discrepancy was resolved by consensus.
A filling defect was depicted as a low attenuating oval or round lesion that represented incomplete mixing of the contrast agent and blood. All filling defects depicted on late-phase images also demonstrated filling defects on early-phase images and were considered “persistent filling defects.” The findings from MDCT were classified into 3 categories: (1) no filling defect; (2) an early filling defect (a filling defect seen on an early-phase image without considering the late-phase images); and (3) a persistent filling defect (a filling defect that was continuously noted on late-phase and early-phase images).
From the multidector computed tomographic findings, several hypotheses were considered if only early-phase imaging were performed or if 2-phase imaging were performed. When we obtained only early-phase images, the hypotheses were as follows: hypothesis 1, an early filling defect is considered a thrombus; hypothesis 2, an early filling defect is considered a thrombus and SEC; and hypothesis 3, an early filling defect is considered a thrombus and severe SEC. When we obtained 2-phase images, we considered the following additional hypotheses: hypothesis 4, an early filling defect depicted only on early-phase images and showing complete homogenous enhancement on late-phase images is considered SEC; hypothesis 5, a persistent filling defect is considered a thrombus; and hypothesis 6, a persistent filling defect is considered a thrombus and severe SEC.
To assess whether quantitative measurement of a filling defect was possible, a reader unaware of the other data calculated the ratio of Hounsfield units in the LAA to the ascending aorta (LAA/AA). To identify the lowest overall Hounsfield unit density in the LAA, the 1-cm 2 circular region of interest was sampled in multiple areas within an axial slice from the LAA apex to the LAA ostium. In addition, a 1-cm 2 region of interest was placed within the contrast-enhanced lumen of the ascending aorta on the same axial plane. The LAA/AA ratio was independently examined on early-phase (LAA/AA E ) and late-phase (LAA/AA L ) images.
Two experienced observers (SJL and HJC) who were unaware of the other data reviewed all transesophageal echocardiographic images. Differences in the assessments by the observers were resolved by consensus.
A thrombus was defined as a circumscribed echogenic or echolucent mass distinct from the surrounding atrial wall. SEC was defined as an intracavitary swirling smoke-like echo within the LA or LAA that could not be eliminated by altering the gain settings. We classified the transesophageal echocardiographic images using a 5-grade scale: (1) none, absence of SEC; (2) mild SEC, minimal echogenicity that was only transiently detectable with optimal gain settings during the cardiac cycle; (3) moderate SEC, a dense swirling pattern of echoes seen during the entire cardiac cycle; (4) severe SEC, intense echogenicity, and a slow swirling pattern in the LAA, usually seen with similar echogenicity in the main cavity; and (5) the presence of a thrombus as a well-contoured echogenic mass.
With TEE as the reference standard, the diagnostic performance of MDCT was assessed using McNemar’s chi-square test. One-way analysis of variance with Scheffe’s post hoc test was used to evaluate the statistical significance of the differences in each of the mean LAA/AA E and LAA/AA L ratios for the 5 grades using TEE. To assess the optimal cutoff value of the LAA/AA ratio between a thrombus and SEC, receiver operating characteristic curves were generated and the area under the curve was calculated. The Statistical Package for Social Sciences for Windows, version 15.0 (SPSS, Chicago, Illinois), was used for the statistical analyses. MedCalc software, version 9.20 (MedCalc Software, Mariakerke, Belgium), was also used for the receiver operating characteristic curve analysis. p Values <0.05 were considered significant.
Intraobserver and interobserver agreement for identifying thrombus using TEE and MDCT were calculated using κ statistics.
Results
The clinical characteristics of the 314 patients in the present study were summarized in Table 1 . Of the 314 patients, 72 (22.9%) had had atrial fibrillation at the time of the stroke, 36 (11.5%) had had valve dysfunction of at least moderate regurgitation in any valve, and 9 (2.9%) had also had valvular dysfunction in >2 different valves.
| Demographic | Value |
|---|---|
| Age (mean, years) | 65.2 ± 12.9 |
| Men | 186 (59.2%) |
| Atrial fibrillation | 72 (22.9%) |
| Hypertension | 154 (49.0%) |
| Diabetes mellitus | 102 (32.5%) |
| Hyperlipidemia | 59 (18.8%) |
| Smoker | 96 (30.6%) |
| Ejection fraction ≤55% | 52 (16.6%) |
| Valve dysfunction | 36 (11.5%) |
| At least moderate mitral regurgitation (≥2+) | 8 (2.5%) |
| At least moderate aortic regurgitation (≥2+) | 11 (3.5%) |
| At least moderate tricuspid regurgitation (≥2+) | 8 (2.5%) |
| At least moderate mitral regurgitation (≥2+) and aortic regurgitation (≥2+) | 2 (0.6%) |
| At least moderate mitral regurgitation (≥2+) and tricuspid regurgitation (≥2+) | 4 (1.3%) |
| At least moderate aortic regurgitation (≥2+) and tricuspid regurgitation (≥2+) | 3 (1.0%) |
TEE demonstrated SEC in 104 patients (33.1%) and the presence of a thrombus in 23 (7.3%). As depicted on the multidetector computed tomographic images, 119 patients (37.9%) demonstrated early filling defects at the LAA on the early-phase images. Of these patients, 29 (9.2%) had a persistent filling defect in the LAA on the added late-phase images. These visual analyses for TEE and MDCT are summarized in Table 2 . The quality of all the multidetector computed tomographic images was considered acceptable for the evaluation of the intracardiac chamber, including the LAA.
| MDCT | TEE | ||||
|---|---|---|---|---|---|
| None (n = 187) | SEC (n = 104) | Thrombus (n = 23) | |||
| Mild (n = 33) | Moderate (n = 44) | Severe (n = 27) | |||
| Early-phase | |||||
| No filling defect (n = 195) | 185 | 8 | 2 | 0 | 0 |
| Early filling defect (n = 119) | 2 | 25 | 42 | 27 | 23 |
| Late-phase | |||||
| Persistent filling defect (n = 29) | 0 | 0 | 0 | 6 | 23 |
From the multidetector computed tomographic findings, we evaluated the diagnostic accuracy of MDCT for each hypothesis, with TEE as the reference standard ( Table 3 ). If an early filling defect at the LAA was considered a thrombus (hypothesis 1), the overall sensitivity, specificity, positive predictive value, and negative predictive value for detection of the thrombus, with TEE as the reference standard, was 100%, 67.0%, 19.3%, and 100%, respectively. However, if an early filling defect was considered a thrombus, including SEC (hypothesis 2), the specificity and positive predictive value were markedly increased at 98.9% and 98.3%, respectively, but the sensitivity and negative predictive value were slightly decreased (92.1% and 94.9%, respectively). In addition, if an early filling defect was considered a thrombus and severe SEC (hypothesis 3), the sensitivity and negative predictive value were both 100%, but the specificity (73.9%) and positive predictive value (42.0%) were slightly increased compared to the values for consideration of hypothesis 1. When considering added late-phase imaging, if a filling defect seen only on early-phase (and was normally enhanced on late-phase) images were considered SEC (hypothesis 4), the overall sensitivity, specificity, positive predictive value, and negative predictive value was 84.6%, 99.1%, 97.8%, and 92.1%, respectively. If a persistent filling defect was considered a thrombus (hypothesis 5), the overall sensitivity, specificity, positive predictive value, and negative predictive value was 100%, 97.9%, 79.3%, and 100%, respectively. Furthermore, if a persistent filling defect was considered a thrombus and severe SEC (hypothesis 6), the overall sensitivity, specificity, positive predictive value, and negative predictive value were all 100%. Therefore, most filling defects that were seen only on early-phase images and that showed complete homogenous enhancement on late-phase images were SEC, with a high sensitivity and specificity ( Figure 1 ). For a persistent filling defect seen on added late-phase images, the presence of a thrombus, as well as severe SEC, should be considered.
| Hypothesis | Sensitivity | Specificity | Positive Predictive Value | Negative Predictive Value |
|---|---|---|---|---|
| When seen only on early-phase images | ||||
| 23/23 (100%) | 195/291 (67.0%) | 23/119 (19.3%) | 195/195 (100%) |
| 95% Confidence interval | 90.852–1.000 | 0.613–0.724 | 0.127–0.275 | 0.981–1.000 |
| 117/127 (92.1%) | 185/291 (98.9%) | 117/119 (98.3%) | 185/195 (94.9%) |
| 95% Confidence interval | 0.860–0.962 | 0.962–0.999 | 0.941–0.998 | 0.908–0.975 |
| 50/50 (100%) | 195/264 (73.9%) | 50/119 (42.0%) | 195/195 (100%) |
| 95% Confidence interval | 0.929–1.000 | 0.682–0.791 | 0.330–0.515 | 0.981–1.000 |
| When seen on added late-phase images | ||||
| 88/104 (84.6%) | 208/210 (99.1%) | 88/90 (97.8%) | 208/224 (92.9%) |
| 95% Confidence interval | 0.762–0.910 | 0.967–0.999 | 0.922–0.997 | 0.887–0.959 |
| 23/23 (100%) | 285/291 (97.9%) | 23/29 (79.3%) | 285/285 (100%) |
| 95% Confidence interval | 0.852–1.000 | 0.956–0.992 | 0.603–0.920 | 0.987–1.000 |
| 29/29 (100%) | 285/285 (100%) | 29/29 (100%) | 285/285 (100%) |
| 95% Confidence interval | 0.881–1.000 | 0.987–1.000 | 0.881–1.000 | 0.987–1.000 |
