Simplified Assessment of Lesion Water Uptake for Identification of Patients within 4.5 Hours of Stroke Onset: An Analysis of the MissPerfeCT Study

Article information

J Stroke. 2022;24(3):390-395

Publication date (electronic) : 2022 September 30

doi : https://doi.org/10.5853/jos.2022.00220

Peter B. Sporns^,a,b,*, Marco Höhne^c,*, Lennart Meyer^d, Christos Krogias^e, Volker Puetz^f,g, Kolja M. Thierfelder^h, Marco Duering^i,j, Daniel Kaiser^g,k, Sönke Langner^h, Alex Brehm^a, Lukas T. Rotkopf^l, Wolfgang G. Kunz^m, Jens Fiehler^b, Walter Heindel^c, Peter Schrammⁿ, Heinz Wiendl^d, Heike Minnerup^o, Marios Nikos Psychogios^a, André Kemmling^p,*, Jens Minnerup^d,*

^aDepartment of Neuroradiology, Clinic of Radiology & Nuclear Medicine, University Hospital Basel, Basel, Switzerland

^bDepartment of Diagnostic and Interventional Neuroradiology, University Medical Center Hamburg-Eppendorf, Hamburg, Germany

^cDepartment of Radiology, University Hospital of Münster, Westfaelische Wilhelms-University of Münster, Münster, Germany

^dDepartment of Neurology with Institute of Translational Neurology, University of Münster, Münster, Germany

^eDepartment of Neurology, St. Josef-Hospital, Ruhr University Bochum, Bochum, Germany

^fDepartment of Neurology, University Hospital Carl Gustav Carus, Dresden, Germany

^gNeurovascular Center Dresden, University Hospital Carl Gustav Carus, Dresden, Germany

^hDepartment of Radiology and Institute of Diagnostic and Interventional Radiology, University Medical Center Rostock, Rostock, Germany

ⁱMedical Image Analysis Center (MIAC) and Department of Biomedical Engineering, University of Basel, Basel, Switzerland

^jInstitute for Stroke and Dementia Research (ISD), University Hospital, LMU Munich, Munich, Germany

^kInstitute of Neuroradiology, University Hospital Carl Gustav Carus, Dresden, Germany

^lDepartment of Radiology, German Cancer Research Center, Heidelberg, Germany

^mDepartment of Radiology, University Hospital, LMU Munich, Munich, Germany

ⁿDepartment of Neuroradiology, University Medical Center Schleswig-Holstein, Lübeck, Germany

^oInstitute of Epidemiology and Social Medicine, University of Münster, Münster, Germany

^pDepartment of Diagnostic and Interventional Neuroradiology, University Medical Center Marburg, Marburg University, Marburg, Germany

Correspondence: Peter B. Sporns Department of Neuroradiology, Clinic of Radiology & Nuclear Medicine, University Hospital Basel, Petersgraben 4, 4031 Basel, Switzerland Tel: 41-61-328-6370 Fax: 41-61-328-6371 E-mail: peter.sporns@hotmail.de

*These authors contributed equally to the manuscript.

Received 2022 January 17; Revised 2022 February 22; Accepted 2022 April 1.

Abstract

Introduction

Current guidelines recommend thrombolysis with intravenous alteplase (IV-tPA) for acute ischemic stroke up to 4.5 hours from stroke onset [1]. Following these guidelines patients with unknown time of symptom onset, representing up to 25% of all stroke patients, are excluded from IV-tPA treatment [2]. There are currently two imaging approaches to identify those patients likely to benefit from IV-tPA treatment among those “wake-up stroke” patients. The first approach uses the identification of the potentially salvageable penumbra by perfusion imaging [3]. The second approach quantifies the mismatch between diffusion weighted imaging (DWI) and fluid attenuated inversion recovery (FLAIR) sequences on magnetic resonance imaging (MRI) to identify those patients with a symptom onset of less than 4.5 hours [4]. Both imaging methods have crucial limitations; the first approach requires the application of dedicated software and is reported to identify a lower proportion of patients compared to the “tissue-clock” approach [5]. The second approach is limited by the restricted availability and increased logistic effort of MRI for acute stroke triage.

Recently, we have established a computed tomography (CT) based method for the quantification of lesion water uptake on admission that is able to identify patients within the 4.5-hour-time window with high accuracy [6]. The aim of this study was to develop a simple method that identifies patients within the 4.5-hour-time window based on simple region of interest (ROI) measurements as a proxy of lesion water uptake in admission CT. This approach is easily and rapidly applicable with standard radiological software worldwide and without requiring additional software tools. We therefore performed a multicenter cohort study including patients with known symptom onset and hypothesized that our ROI-based method would be able to identify those patients in the 4.5-hour-time window eligible for thrombolysis.

Methods

Image analysis

The imaging biomarker for lesion age, the water uptake (w), is based on ischemic and preischemic CT density and was calculated according to the following formula:

w=ΔVwaterVischemic =1−DischemicDpreischemic 

as previously published [6]. The algorithm to classify patients within the 4.5-hour-time window was based on placement of a ROI in the area of the NCCT that corresponded with the region with the lowest CBV. The ROI was chosen as large as possible but at least 1 cm2. As a reference a ROI of the same size was placed in the contralateral corresponding region in the NCCT and the ratio of the affected/unaffected side was calculated. Placement of the ROIs was done independently by two experienced readers (P.B.S. and M.H.) blinded to clinical information (Figure 1 illustrates the method). When judgements were conflicting, images were rated by a third reader (J.M.) and then discussed until consensus was reached between all observers on the optimal ROI placement. The consensus judgement was used for further analysis.

Figure 1.

Illustration of the method of quantification of lesion water uptake for a patients with a symptom onset <4.5 and >4.5 hours. (A) Cerebral blood volume (CBV)-map (mL/100 g) showing low CBV lesion in the right posterior cerebral artery territory. (B) A region of interest (yellow circles) is placed in the area that corresponds to the lowest CBV lesion in the non-contrast computed tomography (NCCT). Mean computed tomography (CT) attenuation in of the lesion on the ischemic side (41.7 Hounsfield unit [HU]) is divided through mean attenuation on the contralateral side (42.1 HU) indicating a quotient of 0.99 which correctly determined the patient to be inside the 4.5-hour-time window. (C) CBV-map (mL/100 g) showing low CBV lesion in the right temporal middle cerebral artery territory. (D) A region of interest (yellow circles) is placed in the area that corresponds to the lowest CBV lesion in the NCCT. Mean CT attenuation of the lesion on the ischemic side (25.9 HU) is divided through mean attenuation on the contralateral side (36.8 HU) indicating a quotient of 0.70 which correctly determined the patient to be outside the 4.5-hour-time window.

Results

Patient baseline characteristics

A total of 263 patients were included in our study of whom 204 (77.6%) presented within and 59 (22.4%) patients later than 4.5 hours of symptom onset. Table 1 shows their baseline characteristics. Groups for patients presenting <4.5 and >4.5 hours were comparable regarding age (mean±standard deviation, 72.3±13.3 years vs. 68.6±13.2 years, P=0.065), gender (women, 52.0% vs. 50.8, P=0.526), and National Institutes of Health Stroke Scale score on admission (median 12.0 [interquartile range, IQR, 9.8 to 14.2] vs. 12.0 [IQR, 9.5 to 14.4], P=0.416). Comorbidities like hypertension (80.3% vs. 73.2%, P=0.252), diabetes mellitus (23.7% vs. 26.8%, P=0.639), and atrial fibrillation (51.0% vs. 39.3%, P=0.121) were also not significantly different between the groups.

Table 1.

Baseline characteristics of patients

Receiver operating characteristic analysis

Median ratio of Hounsfield units (HU) for patients ≤4.5 hours was 6.7% (95% CI, 6.0% to 7.4%) and 12.7% (95% CI, 10.7% to 14.7%) for patients with an onset of >4.5 hours. The AUC discriminating patients within from those beyond 4.5 hours of symptom onset according to percentage of water uptake was 0.744 (Figure 2). Interobserver agreement for ROI placement was 90.9 % with a κ of 0.807 (95% CI, 0.701 to 0.913) indicating an ‘almost perfect’ agreement. The ideal cut-off value for achieving the highest positive predictive value of 88.8% in our study was 0.095 (9.5% water uptake) leading to a sensitivity of 73.5%, a specificity of 67.8%, and a negative predictive value of 42.6%, respectively (Figure 3).

Figure 2.

Receiver operating characteristic (ROC)-curve for identification of patients within 4.5 hours of symptom onset. AUC, area under the curve.

Figure 3.

Cut-off 0.095 for identification of patients within 4.5 hours of symptom onset. PPV, positive predictive value; NPV, negative predictive value.

Discussion

This multicenter study shows that a simple ROI-based assessment of admission CTs identifies patients within 4.5 hours of symptom onset, thus eligible for thrombolysis. As discussed in the DWI-FLAIR mismatch for the identification of patients with acute ischemic stroke within 4.5 hours of symptom onset (PREFLAIR) study, the observational study preceding the randomized Efficacy and Safety of MRI-Based Thrombolysis in Wake-Up Stroke (WAKE-UP) trial, the positive predictive value is likely to be most relevant measure for clinical practice [8]. Thus, the ideal cut-off with a maximum positive predictive value of 88.6% would be 0.095 meaning a water uptake of 9.5% in our study. We have recently shown that the CT-based quantification of brain tissue water uptake identifies stroke patients with symptom onset within 4.5 hours with high accuracy [6]. The pathophysiological basis for this observation is that tissue water uptake after cerebral artery occlusion follows a characteristic course that is visualized by a decreasing CT density [9-11]. However, in our previous study we used specific post-processing software to quantify water uptake within early infarct lesion that is not widely available in routine practice. We therefore developed the concept of a ROI-based assessment that translates the pathophysiological approach of water uptake quantification into a clinically feasible and fast applicable algorithm not requiring a specific software.

Until recently, imaging strategies for determining the time of symptom onset were primarily based on MRI, specifically, the DWI-FLAIR mismatch concept [8]. Following this concept patients with visible changes on DWI but normal FLAIR are likely within the time window of thrombolysis. The multicenter, randomized, double-blind WAKE-UP trial showed that thrombolysis in patients with unknown time of symptom onset guided by MRI DWI-FLAIR mismatch results in a significantly better functional outcome [4]. Overall, the WAKE-UP trial confirms the rationale of the “tissue-clock” approach for the identification of stroke patients with unknown time of symptom onset eligible for thrombolysis. However, compared with MRI, CT is less affected by contraindications (like pacemakers) and is more generally available in the acute setting in most hospitals that treat acute ischemic stroke patients and thus remains the primary imaging modality worldwide.

An alternative approach to identify patients who may benefit from thrombolysis despite unknown time of symptom onset is based on imaging a favorable penumbral pattern in MRI or CT with infarct core-perfusion mismatch as a surrogate of salvageable tissue [3]. A recent systematic review and meta-analysis showed that patients with a favorable penumbral pattern operationally defined by automated software in core laboratory analysis had a significant benefit from thrombolysis [12]. However, automated imaging software required for perfusion mismatch quantification used in clinical studies limits its use in clinical practice. In addition, a recent MRI study found that the yield of a “tissue-clock” imaging approach to select patients eligible for thrombolysis in the unknown time window was double that of infarct core-perfusion mismatch based patient selection [5]. In this context, our ROI-based method has several advantages over existing imaging based methods [13] for the identification of patients with unknown time of symptom onset eligible for thrombolysis including the speedy accessibility of CT worldwide compared to MRI and the dispensability of automated software tools.

A limitation of our study is the retrospective design. However, all images were assessed by readers masked to clinical information. Moreover, the cut-off in this study was not validated in an external cohort. Further, the use of different perfusion imaging processing softwares may lead to distinct findings regarding ischemic volumes on CTP maps [14]. A strength of our study is the almost perfect interobserver agreement for ROI placement that was much better than the interobserver agreement reported for DWI-FLAIR mismatch detection on MRI [8].

Conclusions

In conclusion, ischemic stroke patients with an unknown time of symptom onset can be identified as being within 4.5 hours after symptom onset using a simple ROI-based method of assessment of admission CT. This can potentially guide the decision to administration of thrombolysis in patients with unknown time of stroke onset but the findings need to be validated by prospective, large-scale studies.

Article information Continued