Overestimation of Susceptibility Vessel Sign
A Predictive Marker of Stroke Cause
Background and Purpose—The extent of blooming artifact may reflect the amount of paramagnetic material. We thus assessed the overestimation ratio of susceptibility vessel sign (SVS) on susceptibility-weighted imaging, defined as the extent of SVS width beyond the lumen and examined its value for predicting the stroke cause in acute ischemic stroke patients.
Methods—We included consecutive acute ischemic stroke patients with proximal large artery occlusion who underwent both susceptibility-weighted imaging and time-of-flight magnetic resonance angiography within 8 hours poststroke onset. We calculated the length, width, and overestimation ratio of SVS on susceptibility-weighted imaging and then investigated their values for predicting the stroke cause, respectively.
Results—One-hundred eleven consecutive patients (72 female; mean age, 66.6±13.4 years) were enrolled, among whom 39 (35.1%) were diagnosed with cardiogenic embolism, 43 (38.7%) with large artery atherosclerosis, and 29 (26.1%) with undetermined cause. The presence, length, width, and overestimation ratio of SVS were all independently associated with the cause of cardiogenic embolism after adjusting for baseline National Institute of Health Stroke Scale and infarct volume. After excluded patients with undetermined cause, the sensitivity and specificity of overestimation ratio of SVS for cardiogenic embolism were 0.971 and 0.913; for the length of SVS, they were 0.629 and 0.739; for the width of SVS, they were 0.829 and 0.826, respectively.
Conclusions—The overestimation ratio of SVS can predict cardiogenic embolism, with both high sensitivity and specificity, which can be helpful for the management of acute ischemic stroke patients in hyperacute stage.
The robust data supporting endovascular therapy for acute ischemic stroke (AIS) now place considerable emphasis on noninvasive imaging for the identification of proximal arterial occlusion.1 Cardiogenic embolism (CE) and large artery atherosclerosis (LAA) are major factors in large artery occlusion. Although previous studies have reported on some imaging markers of stroke cause in AIS patients, their occurrence rate, specificity, or sensitivity are not fully characterized.
The composition of clots in obstructed arteries varies depending on whether the embolic source is CE or LAA.2 Susceptibility vessel sign (SVS), defined as the presence of hypointensity in artery with a blooming artifact on susceptibility-weighted imaging (SWI), is useful because it yields information about the site of arterial obstruction.3,4 A previous imaging–pathology correlation study showed that the thrombus component within SVS was predominantly red blood cells in cardiogenic thrombi, suggesting that the presence of SVS might have a predictive effect on CE.2 Recently, based on the observation that the size of blooming artifact may reflect the amount of paramagnetic materials, we introduced a definition of overestimation ratio of SVS, which was calculated with the extent of SVS width beyond the lumen.5 We also found that many CE patients with large artery occlusion manifested large overestimation ratio of SVS. We hypothesized that high overestimation ratio of SVS on SWI in AIS patients might indicate the cause of CE and thus examined its predictive value for stroke cause.
Subjects and Methods
This study was approved by the human ethics committee of our center. Our prospectively collected database was reviewed for patients with AIS who were admitted in our hospital between March 2009 and July 2016. Patients were enrolled who (1) had a diagnosis of ischemic stroke confirmed by diffusion-weighted imaging, (2) underwent SWI and time-of-flight magnetic resonance angiography within 8 hours poststroke onset, and (3) had a proximal arterial occlusion of Willis circle on baseline time-of-flight magnetic resonance angiography. Patients who had poor-image quality because of motion artifacts were excluded.
Magnetic resonance imaging parameters and statistical analysis methods are provided in the online-only Data Supplement.
SWI magnitude images were coregistered and merged to axial source images of time-of-flight magnetic resonance angiography. The length, largest width of SVS, and the width of arteries at the interface on the merged image were measured independently by 2 neurologists who were blinded to the clinical data. The specific measurement methods were the same as previously described.5 SVS was defined as the presence of hypointensity in arteries with a blooming artifact. Thus, the extent of the blooming artifact was expressed as the overestimation of thrombus width: overestimation ratio=width of SVS/width of large artery. The stroke causes were determined according to the TOAST (Trial of ORG 10172 in Acute Stroke Treatment) standard. Receiver operating characteristic curve analysis was used to determine predictive value for CE. We compared the diagnostic odds ratio and area under the curve of each measurement of SVS.
A total of 111 patients were included in analysis (72 female; mean age, 66.6±13.4 years) after exclusion of 4 patients with poor-image quality. Among them, 39 (35.1%) patients were in the CE group, 43 (38.7%) were in the LAA group, and 29 (26.1%) were in the undetermined cause group. Table I in the online-only Data Supplement shows the clinical and image characteristics of the 3 groups. Table II in the online-only Data Supplement shows the detailed TOAST classifications of the 3 stroke causes.
SVS occurred in 80 (72.1%) patients on initial SWI, among whom the median length of the SVS was 10.00 mm (interquartile range, 5.59–15.89 mm), the median width of the SVS was 3.77 mm (interquartile range, 3.13–4.83 mm), and the median overestimation ratio of SVS was 2.08 (interquartile range, 1.57–2.35). The interobserver and intraobserver reliabilities about the measurements were excellent, with intraclass correlation coefficient of 0.989 and 0.975 for SVS length, 0.988 and 0.971 for SVS width, 0.984 and 0.967 for magnetic resonance angiography width of artery.
We retrospectively analyzed 82 patients with determined cause (39 patients in CE group and 43 patients in LAA group). The Table shows the diagnostic odds ratio and area under the curve of each measurement of SVS after adjusting for baseline National Institute of Health Stroke Scale and baseline infarct volume. The optimum cutoff values for predicting CE were 10.16 mm for length of SVS, 3.65 mm for width of SVS, and 2.003 for overestimation ratio of SVS, respectively. An overestimation ratio of SVS >2.003 has the highest sensitivity (97.1%) and specificity (91.3%) for predicting CE among the 3 SVS characteristics. Thirty-six patients had overestimation ratio of SVS >2.003, and 34 (94.4%) were finally diagnosed with CE. The receiver operating characteristic analysis comparing CE, LAA, and stroke of unknown cause was shown in Table III in the online-only Data Supplement. Figure shows the representative characteristics of SVS in patients with different etiologic diagnoses.
Length of SVS, width of SVS, and overestimation rate of SVS were not correlated with onset to imaging time (length of SVS: r=0.026, P=0.828; width of SVS: r=0.095, P=0.423; overestimation rate of SVS: r=0.074, P=0.533) in all patients. Also, these characteristics were not correlated with onset to imaging time in CE group, LAA group, and undetermined cause group, respectively. The results were shown in Figure I in the online-only Data Supplement.
This novel study established that the SVS overestimation ratio is useful in predicting cardioembolism, noting that the SVS on SWI was detected in 89.7% of patients with CE, and the overestimation ratio of SVS could predict CE, with both high sensitivity and specificity of 0.971 and 0.913, respectively.
The appearance of SVS depends on the size of clots and the paramagnetic material content within the clots. The size of clots from CE is thought to be larger than that from LAA.6 This may contribute to the phenomenon that the presence of SVS itself, the length, and width of SVS also had predictive values for CE.
Previous pathological study demonstrated that thrombus originating from CE had higher proportion of red blood cells, compared with noncardiac clot.2 With increasing age of the thrombus, oxyhemoglobin in erythrocytes goes through sequential stages of degradation into deoxyhemoglobin, methemoglobin, and then hemosiderin.7 Deoxyhemoglobin and hemosiderin in the erythrocytes are paramagnetic materials that result in a signal reduction on SWI. The blooming effect of clot therefore reflects the amount of paramagnetic materials within clots. Moreover, different from the clot because of CE, the width of clot in LAA patients with intracranial atherosclerotic disease should be smaller than the diameter of embolic blood vessel6 because the formation of clot is superimposed on the atheromatous plaque subtending the vessel lumen. This would lead to a limited blooming artifact along the width of clot in LAA, provided with similar clot materials.
Most recently, multiple endovascular stroke trials have demonstrated that reperfusion within 8 hours after stroke onset can accomplish high reperfusion rate, thus improve functional outcome.8,9 Different devices and different combined therapies have been introduced to treat AIS with different causes.10 Thus, the predictive value of overestimation ratio of SVS for CE in hyperacute phase would be helpful to the decision-making of reperfusion therapy. Besides, early prediction of AIS can bring forward secondary prevention strategy, which is also of great importance in AIS treatment. In addition, the time profile analysis indicated that the SVS characteristics did not change with the onset to imaging time, which also demonstrated the validity of overestimation ratio of SVS in predicting CE in hyperacute phase of AIS.
Limitations include the retrospective nature in a single stroke center and relatively small number of patients. This might have a potential risk of selection bias because patients were predominantly female, relatively young, and with relatively small infarct volumes in our cohort, although data were prospectively established using a stroke registry and magnetic resonance imaging protocol. The high morbidity of intracranial large arterial atherosclerotic stroke has a large influence on the magnetic resonance imaging profile and severity, which may explain that National Institute of Health Stroke Scale and infarct volume were relatively small in some patients with LAA group in our cohort. Second, the results in this study depend on the magnetic resonance imaging parameters described above because the extent of the blooming artifact might vary with different magnetic field strengths and echo time. Third, the hypothesis on clot compositions lacked pathological basis, which needs further investigation. Final, if the size of thrombus mainly contributes to SVS overestimation, clean intracranial vessels in LAA population would reduce the predictive power of SVS overestimation. Because Asian population has a high risk of intracranial atherosclerotic disease, whereas Western population has more carotid disease, the results need to be confirmed from other centers especially in Western population.
The overestimation ratio of SVS can predict CE, with a higher sensitivity and specificity over the other SVS characteristics. As an image marker available in hyperacute phase, SVS overestimation can provide cause information other than convention examinations, which may be helpful to decision of reperfusion therapy in hyperacute phase and bring forward the secondary prevention strategy.
Sources of Funding
This work was supported by the National Natural Science Foundation of China (81471170 and 81622017) and the National Key Research and Development Program of China (2016YFC1301500).
Dr Liebeskind is a consultant/advisory board (modest) for Stryker and Medtronic. The other authors report no conflicts.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.117.016727/-/DC1.
- Received January 14, 2017.
- Revision received March 28, 2017.
- Accepted April 3, 2017.
- © 2017 American Heart Association, Inc.
- Kim SK,
- Yoon W,
- Kim TS,
- Kim HS,
- Heo TW,
- Park MS
- Flacke S,
- Urbach H,
- Keller E,
- Träber F,
- Hartmann A,
- Textor J,
- et al
- Yan S,
- Chen Q,
- Zhang X,
- Xu M,
- Han Q,
- Shao A,
- et al
- Behme D,
- Molina CA,
- Selim MH,
- Ribo M