Abstract
Background and Purpose
Ischemic stroke patients with regional very low cerebral blood volume (VLCBV) on baseline imaging have increased risk of parenchymal hemorrhage (PH) following intravenous alteplase-induced reperfusion. We developed a method for automated detection of VLCBV and examined whether patients with reperfused-VLCBV are at increased risk of PH following endovascular reperfusion therapy.
Methods
Receiver-operating-characteristic (ROC) analysis was performed to optimize a relative cerebral blood volume (rCBV) threshold associated with PH in patients from the DEFUSE 2 study. Regional reperfused-VLCBV was defined as regions with low rCBV on baseline imaging that demonstrated normal perfusion (Tmax<6s) on coregistered early follow-up MRI. The association between VLCBV, regional reperfused-VLCBV and PH was assessed in univariate and multivariate analyses.
Results
In 91 patients, the greatest area-under-curve (AUC) for predicting PH occurred at an rCBV threshold of <0.42 (AUC 0.77). At this threshold, VLCBV lesion volume ≥3.55 ml optimally predicted PH with 94% sensitivity and 63% specificity. Reperfused-VLCBV lesion volume was more specific (0.74) and equally sensitive (0.94). In total 18 patients developed PH, of whom 17 presented with VLCBV (39% vs 2 %; p=0.001), all of them had regional reperfusion (47% vs 0%, p=0.01) and 71% received IV alteplase. VLCBV lesion (OR 33) and bridging with IV alteplase (OR 3.8) were independently associated with PH. In a separate model, reperfused-VLCBV remained the single independent predictor of PH (OR 53).
Conclusions
These results suggest that VLCBV can be used for risk stratification of patients scheduled to undergo endovascular therapy in trials and routine clinical practice.
Keywords: Perfusion imaging, Computed tomography, endovascular treatment, hemorrhage, low CBV
Introduction
Parenchymal hematoma (PH) is the most feared complication of reperfusion therapy in acute ischemic stroke. Imaging characteristics that are associated with an increased risk of parenchymal hematoma include a large DWI lesion1, a lesion with a severely prolonged Tmax2, 3, a very low ADC4, or a very low cerebral blood volume (VLCBV)5, 6. Amongst these variables, VLCBV appears to be the best predictor with very high sensitivity and moderate specificity for predicting PH after intravenous thrombolysis.5,7 Previous studies investigating the association of VLCBV with PH involved manual processing to obtain VLCBV measurements and were based on data from patients treated with intravenous alteplase (iv tPA).5–7 Here, we evaluate if patients with VLCBV can be identified with automated image processing software, and if the presence of VLCBV is associated with the development of parenchymal hemorrhage following endovascular reperfusion.
Methods
Patients
The Diffusion and Perfusion Imaging Evaluation for Understanding Stroke Evolution 2 (DEFUSE 2) study was a prospective observational study of patients who were treated with endovascular therapy.8 The eligibility criteria for the DEFUSE 2 study were intention to start endovascular stroke therapy within 12 hours of symptom onset, age ≥18 years, baseline NIHSS ≥5, non-pregnant state, pre-morbid mRS ≤2, and no contraindication for MRI.8
Imaging protocol and analysis
A standardized imaging protocol using 1.5 or 3 Tesla MRI systems was used. Patients received three scans: a baseline MR scan (gradient recalled echo, intracranial magnetic resonance angiogram, diffusion and perfusion sequence obtained within 90 min before the start of the endovascular procedure), an early follow-up scan (same sequences as baseline) within 12h after the endovascular procedure, and a late follow-up scan (gradient recalled echo, diffusion, and fluid-attenuated inversion recovery) on day 5 or at discharge from the hospital, whichever came sooner.8 Additional imaging was obtained as clinically indicated.
Relative Cerebral Blood Volume (rCBV) maps were generated using fully automated image processing software (RAPID).9 Relative CBV values were calculated for each pixel by dividing its CBV by a smoothed CBV of its mirror-pixel in the contralateral hemisphere. For each patient, the rCBV lesion volume was calculated for each rCBV threshold ranging between 0 and 1 in 0.01 increments. Assessment of rCBV lesions was restricted to brain territory with abnormal perfusion (Tmax>6s) to reduce artifact.7 The rCBV ratio threshold and the rCBV lesion volume threshold that were associated with the best prediction of PH were determined with Receiver Operator Curve (ROC) analyses and used to define patients who had very low CBV (VLCBV). More specifically, optimal VLCBV criteria were defined by first establishing the rCBV ratio threshold with the largest area under the ROC (AUC) and next, the rCBV lesion volume, at this rCBV ratio threshold, with the highest Youden’s index.
Regional reperfusion of the VLCBV lesion was assessed by two investigators on the co-registered subacute MR perfusion (MRP) scan, obtained 12 hours after endovascular therapy. It was defined as restoration of perfusion (Tmax <6s) of the baseline VLCBV lesion. Patients with regional reperfusion of their VLCBV lesion were categorized as “regional reperfused-VLCBV”. Reperfusion was also assessed globally by an experienced neuroendovascular radiologist (MM) on the digital subtraction angiography (DSA). Patients with VLCBV and a modified TICI score of 2b-3 were considered to have “global reperfused-VLCBV”.10 Collateral status was rated on the DSA images by the same investigator according a previously defined 5-point system where 0 is no collateral flow and 4 is complete and rapid collateral flow to the ischemic territory.11 Target mismatch pattern on baseline MRI was defined per criteria used in the parent study.8 The presence of PH (either PH1 or PH2) was assessed on the gradient recalled echo (GRE) sequence of the subacute MRI.12–14
Statistical Analysis
Univariate analyses were used to assess the association of clinical and radiological variables with two outcome variables: VLCBV and PH. Independent predictors of PH were assessed with multivariate analyses. Variables that were significant at an alpha of 0.1 in the univariate analyses were entered in a multivariate logistic regression model. A backward elimination procedure was used in which variables with an alpha of >0.05 were eliminated from the model. Separate multivariate models were constructed for VLCBV and reperfused-VLCBV, to avoid the simultaneous inclusion of predictor variables that are closely correlated. For other predictors that are correlated (eg NIHSS and DWI lesion volume), only the predictor that had the strongest association with PH was included in the multivariate model. Differences in distributions of continuous variables were assessed with the Mann Witney U test. For the distribution of the modified Ranking Scale (mRS), category 5 and 6 were collapsed into one category. Analyses were conducted in SAS 9.4 (SAS Institute, Cary, NC) and StatsDirect (UK).
Results
The DEFUSE 2 study reported findings on 99 patients.8 Of these, 91 patients had adequate baseline and subacute MRI data for assessment of VLCBV, regional reperfusion, and PH. These 91 patients were included in this analysis (Figure 1). The optimal rCBV lesion criteria on baseline MRI to predict PH were an rCBV ratio <0.42 (AUC 0.768) and a volume ≥3.55 ml (Youden’s index 0.578). These criteria were implemented in automated perfusion processing software. VLCBV was present on the baseline MRI in 48% (n=44) of the patients. Compared to patients without VLCBV, patients with VLCBV had larger DWI lesions and higher baseline NIHSS scores, while collateral rating and age did not differ between groups (table 1). Among patients with VLCBV, 82% (36 of 44) had regional reperfusion on follow-up PWI and 43% (19/44) had global reperfusion (TICI 2b or 3).
Figure 1. Flow diagram demonstrating risk of parenchymal hematoma.
A flow diagram illustrates the risk of parenchymal hemorrhage stratified by the presence of VLCBV, regional reperfusion, and pretreatment with intravenous tissue plasminogen activator.
Table 1.
Characteristics for groups with and without VLCBV.
| VLCBV present (n=44) | VLCBV absent (n=47) | P† | |
|---|---|---|---|
| Age, median (IQR) | 66 (50–75) | 74 (54–81) | 0.1 |
| NIHSS, median (IQR) | 19 (14–21) | 13 (9–18) | <0.01 |
| DWI lesion volume, median (range) | 31 ml (2–310 ml) | 7 ml (0.5–65 ml) | <0.01 |
| Target mismatch, no. (%) | 32 (43%) | 12 (75%) | 0.02 |
| Poor Collateral Flow*, no. (%) | 21 (62%) | 21 (54%) | 0.5 |
VLCBV, Very Low Cerebral Blood Volume; DWI, Diffusion Weighted Imaging; NIHSS, National Institute of Health Stroke Scale; IQR, Interquartile Range
P values are obtained from univariate logistic regression analysis
Poor Collateral Flow is defined as a score of 0–2 on a 5-point scale.12
Parenchymal hematoma occurred in 39% (17 of 44) of patients with VLCBV compared to 2% (1 of 47) of patients without VLCBV (p<0.0001, figure 1). The test characteristics of VLCBV for predicting PH in the overall cohort and separately in patients with and without the target mismatch are listed in Table 2. This table also reports the hemorrhage rates and test characteristics of VLCBV in the setting of regional and global reperfusion.
Table 2.
Test characteristics of VLCBV to predict parenchymal hematoma.
| Patient group | PH+/VLCBV+ | PH+/VLCBV− | Sensitivity | Specificity | PPV | NPV |
|---|---|---|---|---|---|---|
| VLCBV | 17/44 | 1/47 | 94% (71–100%) | 63% (51–74%) | 39% (25–54%) | 98% (87–100%) |
| VLCBV in TMM | 13/32 | 1/43 | 93% (64–100%) | 69% (56–78%) | 41% (24–59%) | 98% (86–100%) |
| VLCBV in non TMM | 4/12 | 0/4 | 100% (40–100%) | 33% (11–65%) | 33% (11–65%) | 100% (40–100%) |
| Regional reperfused-VLCBV | 17/36 | 1/55 | 94% (71–100%) | 74% (62–83%) | 47% (31–64%) | 98% (89–100%) |
| Global reperfused-VLCBV | 9/19 | 9/72 | 50% (27–73%) | 86% (76–93%) | 47% (25–71%) | 88% (77–94%) |
VLCBV, Very Low Cerebral Blood Volume; PH, Parenchymal Hematoma; TMM, Target Mismatch; PPV, Positive Predictive Value; NPV, Negative Predictive Value. Regional reperfused VLCBV includes patients with VLCBV who have reperfusion of their VLCBV lesion on follow-up PWI. Global reperfused VLCBV includes patients with VLCBV who have TICI 2B-3 scores on their angiogram at completion of the endovascular procedure.
Significant baseline predictors of PH in univariate analyses were the NIHSS score, DWI lesion volume, bridging with IV tPA, and the presence of VLCBV or regional reperfused-VLCBV (table 3). In multivariate analysis, VLCBV remained as an independent predictor of PH (OR 33, 95% CI 4.0 – 270 P<0.001) in addition to bridging with IV tPA (OR 3.8, 95%CI 1.1 – 13.5, p=0.04). In a separate model, regional reperfused-VLCBV remained as the single independent predictor of PH with an OR of 53 (95% CI 6.4–439.7, P<0.001). In this multivariate model bridging with IV tPA was removed in the backward elimination procedure because the association between this variable and PH was borderline significant (OR 3.7, 95% CI 0.97–14.1, p=0.056).
Table 3.
Characteristics of patients with and without parenchymal hematoma.
| Parenchymal Hematoma (n=18) | No Parenchymal Hematoma (n=73) | P† | |||
|---|---|---|---|---|---|
| Clinical Variables | |||||
| Age, y, mean (SD) | 66.1 | (14.3) | 64.5 | (16.2) | 0.7 |
| Baseline NIHSS, median (IQR) | 19 | (14–21) | 14 | (11–19) | 0.02 |
| SBP, mm Hg, mean (SD) | 149.1 | (24.6) | 144.9 | (22.1) | 0.5 |
| Anticoagulant, no. (%) | 2 | (11.1%) | 10 | (13.7%) | 0.8 |
| Antiplatelet, no. (%) | 7 | (38.9%) | 26 | (35.6%) | 0.2 |
| Diabetes Mellitus, no. (%) | 13 | (72.2%) | 4 | (5.5%) | 0.7 |
| IV alteplase, no. (%) | 13 | (72.2%) | 34 | (46.6%) | 0.06 |
| Radiological variables | |||||
| Mean DWI volume, ml (SD) | 54.7 | (75.5) | 21.2 | (22.7) | 0.02 |
| VLCBV, no. (%) | 17 | (94.4%) | 27 | (37.0%) | <0.01 |
| Regional reperfused-VLCBV, no. (%) | 17 | (94.4%) | 19 | (26%) | <0.01 |
| Target mismatch profile, no. (%) | 14 | (78%) | 61 | (84%) | 0.7 |
| Poor collateral flow, no. (%) | 10 | (59%) | 32 | (57%) | 0.9 |
| 90 day functional outcome | |||||
| Median (IQR) mRS | 4 | (1.75–5.25) | 3 | (1–4) | 0.07 |
| mRS 5–6, no. (%) | 8 | (44%) | 13 | (18%) | 0.02 |
NIHSS indicates National Institutes of Health Stroke Scale; SBP, systolic blood pressure; IV, intravenous; DWI, diffusion weighted imaging; VLCBV, very low cerebral blood volume; Poor collateral flow is defined as a score of 0–2 on a 5-point collateral scale.12
P values are obtained from univariate logistic regression analyses
VLCBV was a predictor of poor functional outcome (mRS>2) at 90 days (OR 2.7; 95% CI 1.1–6.2). After adjusting for age and baseline NIHSS the OR for poor functional outcome was 2.9 (95% CI 1–8.5, p = 0.05). VLCBV was also associated with a shift in the distribution of mRS towards worse functional outcome (p=0.047).
Discussion
Our results demonstrate that VLCBV is a strong predictor of parenchymal hematoma in patients undergoing endovascular treatment for acute stroke, particularly in the setting of regional reperfusion. The optimal definition for VLCBV was an rCBV lesion of at least 3.55 mL at an rCBV threshold of <0.42. VLCBV defined with these criteria had excellent negative predictive value and moderate positive predictive value for predicting PH, both in patients with and without the target mismatch pattern. Given those test characteristics, the absence of VLCBV can be used to reassure physicians and patients of a very low risk of PH following endovascular treatment.
In order for an imaging characteristic, such as VLCBV, to be used for risk stratification, it is important to be able to assess its presence in a reliable, reproducible, and consistent manner. Automated image analysis programs can be useful in this regard.15 We integrated the VLCBV algorithm within an existing fully automated post-processing software suite for perfusion imaging (RAPID). This program generated easy to interpret VLCBV maps (figure 2) within 5 minutes of processing time, which permits its use for real-time risk stratification.9
Figure 2. Example of parenchymal hematoma in a patient with VLCBV.
A 50-year old patient presented with a left hemispheric stroke secondary to a left MCA occlusion. The patient’s baseline perfusion scan, processed with a fully automated VLCBV analysis program, demonstrates a 21.15 ml region of VLCBV in the left MCA territory (panel A shows the standard CBV map and panel B shows the CBV map with the VLCBV lesion segmented in purple). Complete reperfusion of this VLCBV lesion was demonstrated on follow-up imaging and a parenchymal hematoma is seen in the region corresponding to the baseline VLCBV lesion on the follow-up GRE sequence (C).
Our methodology for defining VLCBV differs some from prior studies. Specifically, our methodology is based on a very low CBV relative to a mirrored region in the contralateral hemisphere, whereas prior studies used the entire contralateral hemisphere as the reference region. Nevertheless, the results of this study are in line with those of prior studies which also report excellent sensitivity/negative predictive value and moderate to good specificity/positive predictive value for predicting PH based on VLCBV.5–7,16
Our finding that specificity is improved in the presence of regional reperfusion of the VLCBV lesion is also in accordance with prior studies.7 In our study, patients with VLCBV developed PH exclusively in the setting of regional reperfusion (figure 1). In addition to regional reperfusion, we evaluated the impact of global reperfusion (ie reperfusion rated according to post-procedure modified TICI scores) on the association between VLCBV and PH. In the presence of global reperfusion, specificity for predicting PH increased but, due to poor spatial agreement between the region of reperfusion and the VLCBV lesion, sensitivity was dramatically reduced (from 94% to 50%). Taken together with the results of the prior studies, our results indicate that VLCBV is a predictor of PH in the setting of regional reperfusion, regardless of whether reperfusion is achieved through iv thrombolysis or endovascular therapy.
In addition to VLCBV, bridging with IV tPA was an independent predictor of PH following endovascular therapy in our study. This association could not be evaluated in previous VLCBV studies because these studies were limited to patients treated with IV tPA alone.5, 7 The association between bridging with IV tPA and PH was slightly weaker in the multivariate model with reperfused-VLCBV compared to the model with VLCBV, suggesting that the effect of tPA may, in part, be mediated by tPA-induced reperfusion. However, even after adjusting for reperfused-VLCBV, our results show a trend towards an independent association between bridging with IV tPA and PH (p=0.06). This is a novel finding, that suggests that bridging with IV tPA may increase the risk of PH following endovascular reperfusion. These findings should, however, strictly be viewed as hypothesis generating and require validation in other datasets as they are in contrast with prior studies which have not shown an association between bridging with IV tPA and PH or SICH following endovascular therapy.17–20
The VLCBV criteria defined in this study and the test characteristics of VLCBV for predicting PH will also need to be validated. A larger dataset is necessary to confirm that VLCBV improves the risk stratification compared to the sole use of a DWI lesion volume threshold. Given the increasing use of CT perfusion prior to endovascular therapy, future studies should further elucidate whether an analogous approach using CT perfusion can equal the performance of VLCBV based on MR perfusion. Data suggest that imaging prediction of hemorrhage using CT perfusion may require slightly different parameters.2
If the results of this study are validated, VLCBV can be used for risk stratification of patients scheduled to undergo endovascular therapy in trials and routine clinical practice. While the modest positive predictive value precludes its use as a criterion to exclude patients from endovascular therapy, its excellent sensitivity and negative predictive value can reassure physicians and patients of the relative safety of endovascular treatment in the absence of VLCBV.
Acknowledgments
Funding:
DEFUSE 2 was funded by grants from the National Institute for Neurological Disorders and Stroke (R01 NS03932505 to GWA, K23 NS051372 to MGL, and 1ZIANS003043 to SW).
DEFUSE 2 received grant funding from the National Institute of Health (NIH).
BC reports funding of the National Health and Medical Research Council of Australia.
Footnotes
Conflict of Interest:
NKM, AW, BCV, MM, SK, CC, MM, MGL - none.
SC - Consultant for IschemaView Inc
MS - Consultant for IschemaView Inc and minor shareholder of IschemaView Inc
RB - Shareholder of iSchemaView Inc
GWA - Consultant for IschemaView Inc and shareholder of iSchemaView Inc; Advisory Board: Covidien and Lundbeck
Stanford receives royalty payments for RAPID.
References
- 1.Singer OC, Humpich MC, Fiehler J, Albers GW, Lansberg MG, Kastrup A, et al. Risk for symptomatic intracerebral hemorrhage after thrombolysis assessed by diffusion-weighted magnetic resonance imaging. Ann Neurol. 2008;63:52–60. doi: 10.1002/ana.21222. [DOI] [PubMed] [Google Scholar]
- 2.Yassi N, Parsons MW, Christensen S, Sharma G, Bivard A, Donnan GA, et al. Prediction of poststroke hemorrhagic transformation using computed tomography perfusion. Stroke. 2013;44:3039–43. doi: 10.1161/STROKEAHA.113.002396. [DOI] [PubMed] [Google Scholar]
- 3.Kim JH, Bang OY, Liebeskind DS, Ovbiagele B, Kim G-M, Chung CS, et al. Impact of baseline tissue status (diffusion-weighted imaging lesion) versus perfusion status (severity of hypoperfusion) on hemorrhagic transformation. Stroke. 2010;41:e135–42. doi: 10.1161/STROKEAHA.109.563122. [DOI] [PubMed] [Google Scholar]
- 4.Tong DC, Adami A, Moseley ME, Marks MP. Relationship between apparent diffusion coefficient and subsequent hemorrhagic transformation following acute ischemic stroke. Stroke. 2000;31:2378–84. doi: 10.1161/01.str.31.10.2378. [DOI] [PubMed] [Google Scholar]
- 5.Campbell BC, Christensen S, Butcher KS, Gordon I, Parsons MW, Desmond PM, et al. Regional very low cerebral blood volume predicts hemorrhagic transformation better than diffusion-weighted imaging volume and thresholded apparent diffusion coefficient in acute ischemic stroke. Stroke. 2010;41:82–8. doi: 10.1161/STROKEAHA.109.562116. [DOI] [PubMed] [Google Scholar]
- 6.Alsop DC, Makovetskaya E, Kumar S, Selim M, Schlaug G. Markedly reduced apparent blood volume on bolus contrast magnetic resonance imaging as a predictor of hemorrhage after thrombolytic therapy for acute ischemic stroke. Stroke; a journal of cerebral circulation. 2005;36:746–50. doi: 10.1161/01.STR.0000158913.91058.93. [DOI] [PubMed] [Google Scholar]
- 7.Campbell BC, Christensen S, Parsons MW, Churilov L, Desmond PM, Barber PA, et al. Advanced imaging improves prediction of hemorrhage after stroke thrombolysis. Ann Neurol. 2013;73:510–9. doi: 10.1002/ana.23837. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Lansberg MG, Straka M, Kemp S, Mlynash M, Wechsler LR, Wilder MJ, et al. MRI profile and response to endovascular reperfusion after stroke (DEFUSE 2): a prospective cohort study. Lancet Neurol. 2012;11:860–7. doi: 10.1016/S1474-4422(12)70203-X. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Straka M, Albers GW, Bammer R. Real-time diffusion-perfusion mismatch analysis in acute stroke. J Magn Reson Imaging. 2010;32:1024–37. doi: 10.1002/jmri.22338. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Broderick JP, Palesch YY, Demchuk AM, Yeatts SD, Khatri P, Hill MD, et al. Endovascular therapy after intravenous t-PA versus t-PA alone for stroke. N Engl J Med. 2013;368:1265. doi: 10.1056/NEJMoa1214300. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Higashida RT, Furlan AJ, Roberts H, Tomsick T, Connors B, Barr J, et al. Technology Assessment Committee of the American Society of Interventional and Therapeutic Neuroradiology; Technology Assessment Committee of the Society of Interventional Radiology. Trial design and reporting standards for intra-arterial cerebral thrombolysis for acute ischemic stroke. Stroke. 2003;34:e109–e137. doi: 10.1161/01.STR.0000082721.62796.09. [DOI] [PubMed] [Google Scholar]
- 12.Berger C, Fiorelli M, Steiner T, Schäbitz WR, Bluhmki E, Hacke W, et al. Hemorrhagic transformation of ischemic brain tissue: asymptomatic or symptomatic? Stroke. 2001;32:1330–5. doi: 10.1161/01.str.32.6.1330. [DOI] [PubMed] [Google Scholar]
- 13.Fiorelli M, Bastianello S, von Kummer R, del Zoppo GJ, Larrue V, Lesaffre E, et al. Hemorrhagic transformation within 36 hours of a cerebral infarct: relationships with early clinical deterioration and 3-month outcome in the European Cooperative Acute Stroke Study I (ECASS I) cohort. Stroke. 1999;30:2280–4. doi: 10.1161/01.str.30.11.2280. [DOI] [PubMed] [Google Scholar]
- 14.Hacke W, Kaste M, Fieschi C, von Kummer R, Davalos A, Meier D, et al. Randomised double-blind placebo-controlled trial of thrombolytic therapy with intravenous alteplase in acute ischaemic stroke (ECASS II). Second European-Australasian Acute Stroke Study Investigators. Lancet. 1998;352:1245–51. doi: 10.1016/s0140-6736(98)08020-9. [DOI] [PubMed] [Google Scholar]
- 15.Mlynash M, Eyngorn I, Bammer R, Moseley M, Tong DC. Automated method for generating the arterial input function on perfusion-weighted MR imaging: validation in patients with stroke. AJNR American journal of neuroradiology. 2005;26:1479–86. [PMC free article] [PubMed] [Google Scholar]
- 16.Hermitte L, Cho T-H, Ozenne B, Nighoghossian N, Mikkelsen IK, Ribe L, et al. Very low cerebral blood volume predicts parenchymal hematoma in acute ischemic stroke. Stroke. 2013;44:2318–20. doi: 10.1161/STROKEAHA.113.001751. [DOI] [PubMed] [Google Scholar]
- 17.Smith WS, Sung G, Saver J, Budzik R, Duckwiler G, Liebeskind DS, et al. Mechanical thrombectomy for acute ischemic stroke:final results of the Multi MERCI trial. Stroke. 2008;39:1205–12. doi: 10.1161/STROKEAHA.107.497115. [DOI] [PubMed] [Google Scholar]
- 18.Nogueira RG, Gupta R, Jovin TG, Levy EI, Liebeskind DS, Zaidat OO, et al. Predictors and clinical relevance of hemorrhagic transformation after endovascular therapy for anterior circulation large vessel occlusion strokes: a multicenter retrospective analysis of 1122 patients. J Neurointerv Surg. 2015;7:16–21. doi: 10.1136/neurintsurg-2013-010743. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Akins PT, Amar AP, Pakbaz RS, Fields JD SWIFT investigators. Complications of endovascular treatment for acute stroke in the SWIFT trial with solitaire and Merci devices. AJNR. 2014;35:524–8. doi: 10.3174/ajnr.A3707. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Kass-Hout T, Kass-Hout O, Mokin M, Thesier DM, Yashar P, Orion D, et al. Is Bridging with Intravenous Thrombolysis of Any Benefit in Endovascular Therapy for Acute Ischemic Stroke? World Neurosurgery. 2014;82:e453–e458. doi: 10.1016/j.wneu.2013.01.097. [DOI] [PubMed] [Google Scholar]