You are here

  • Home
  • Online First
  • Assessment of Perfusion Volumes by a New Automated Software for Computed Tomography Perfusion

Article Text

Article menu
Original research
Assessment of Perfusion Volumes by a New Automated Software for Computed Tomography Perfusion
  1. Zhixin Cao1,
  2. David Wang2,
  3. Xueyan Feng1,
  4. Pengfei Yang3,4,
  5. Hao Wang5,
  6. Ziqi Xu6,
  7. Yahui Hao7,
  8. Wanxing Ye8,
  9. Fengwei Chen8,
  10. Liyuan Wang1,
  11. Manjun Hao1,
  12. Na Wu1,
  13. Kai-Xuan Yang8,
  14. Yunyun Xiong1,8,9,
  15. Yongjun Wang1,8
  1. 1Department of Neurology, Beijing Tiantan Hospital, Capital Medical University, Beijing, China
  2. 2Neurovascular Division, Department of Neurology, Barrow Neurological Institute, St Joseph's Hospital and Medical Center, Phoenix, Arizona, USA
  3. 3Neurovascular Center, Changhai Hospital, Naval Medical University, Shanghai, China
  4. 4Changhai Clinical Research Unit, Changhai Hospital, Naval Medical University, Shanghai, China
  5. 5Department of Neurology, Linyi People's Hospital, Linyi, Shandong, China
  6. 6Department of Neurology, The First Affiliated Hospital, School of Medicine, Zhejiang University, Hangzhou, Zhejiang, China
  7. 7China National Clinical Research Center for Neurological Diseases, Beijing Tiantan Hospital, Capital Medical University, Beijing, China
  8. 8China National Clinical Research Center for Neurological Diseases, Beijing, China
  9. 9Chinese Institute for Brain Research, Beijing, China
  1. Correspondence to Dr Yunyun Xiong; xiongyunyun{at}bjtth.org; Dr Yongjun Wang; yongjunwang{at}ncrcnd.org.cn

Abstract

Introduction To compare the perfusion volumes assessed by a new automated CT perfusion (CTP) software iStroke with the circular singular value decomposition software RAPID and determine its predictive value for functional outcome in patients with acute ischaemic stroke (AIS) who underwent endovascular treatment (EVT).

Methods Data on patients with AIS were collected from four hospitals in China. All patients received CTP followed by EVT with complete recanalisation within 24 hours of symptom onset. We evaluated the agreement of CTP measures between the two softwares by Spearman’s rank correlation tests and kappa tests. Bland-Altman plots were used to evaluate the agreement of infarct core volume (ICV) on CTP and ground truth on diffusion-weighted imaging (DWI). Logistic regression models were used to test the association between ICV on these two softwares and functional outcomes.

Results Among 326 patients, 228 had DWI examinations and 40 of them had infarct volume >70 mL. In all patients, the infarct core and hypoperfusion volumes on iStroke had a strong correlation with those on RAPID (ρ=0.68 and 0.66, respectively). The agreement of large infarct core (volume >70 mL) was substantial (kappa=0.73, p<0.001) between these two softwares. The ICV measured by iStroke and RAPID was significantly correlated with independent functional outcome at 90 days (p=0.009 and p<0.001, respectively). In patients with DWI examinations and those with an ICV >70 mL, the ICV of iStroke and RAPID was comparable on individual agreement with ground truth.

Conclusion The automatic CTP software iStroke is a reliable tool for assessing infarct core and mismatch volumes, making it clinically useful for selecting patients with AIS for acute reperfusion therapy in the extended time window.

  • Stroke
  • CT perfusion
  • Perfusion
  • infarct core
  • computed tomography perfusion
  • automated software

Data availability statement

Data are available upon reasonable request.

http://creativecommons.org/licenses/by-nc/4.0/

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.

https://doi.org/10.1136/svn-2023-002964

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    WHAT IS ALREADY KNOWN ON THIS TOPIC

    • Automated CT perfusion (CTP) software for evaluating the perfusion volumes is helpful for decision-making for endovascular treatment (EVT) in patients who had a stroke due to large vessel occlusion.

    WHAT THIS STUDY ADDS

    • A new automated CTP software iStroke demonstrates good agreement with the ground truth and the RAPID software for evaluating infarct core and penumbra.

    HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

    • iStroke software can be used for measurement of acute infarct core volume, penumbral tissue and case selection for thrombolysis or EVT in the late time window.

    Introduction

    Current challenges in the management of acute ischaemic stroke (AIS) mainly include optimising imaging analysis methods for patient selection and predicting the clinical outcome before reperfusion therapy.1 The DAWN (Diffusion-Weighted Imaging or Computed Tomography Perfusion Assessment With Clinical Mismatch in the Triage of Wake-Up and Late Presenting Strokes Undergoing Neurointervention With Trevo) and DEFUSE 3 (Diffusion and Perfusion Imaging Evaluation for Understanding Stroke Evolution) trials confirmed that mechanical thrombectomy (MT) improved clinical outcome for patients with AIS selected by either CT perfusion (CTP) or MRI.2 3 Therefore, currently, accurate and reliable measurement of perfusion or diffusion volumes is important for reperfusion therapy decision-making in stroke.

    CTP is widely used to estimate the extent of ischaemic core and penumbra by calculating perfusion parameters including cerebral blood flow (CBF), cerebral blood volume, mean transit time and time-to-maximum (Tmax).4 These perfusion parameters are inferred from CTP images by deconvolution of the concentration–time curves (CTCs)5 against the arterial input function (AIF). Numerous methods have been developed to deconvolve CTC,5 6 and the standard approaches are based on circular singular value decomposition (cSVD).7–9 The RAPID software (iSchema View, Menlo Park, California, USA) is the most popular software based on cSVD algorithm. Several large clinical studies including the DAWN and DEFUSE 3 trials are also supported by the RAPID software.2 3 However, cSVD algorithms are notoriously sensitive to noise and can produce varying results depending on their implementation.10 More recently, numerous attempts have been made to use neural networks for estimating perfusion parameters.11–14 Furthermore, this technique is more robust to noise and scan artefacts compared with the cSVD software. In a previous study, we demonstrated that the automated software iStroke (Beijing Tiantan Hospital and Biomind, Beijing, China) predicted perfusion parameters on magnetic resonance perfusion-weighted imaging.15 However, whether iStroke has similar accuracy by measuring perfusion volumes on CTP using deep neural networks to the RAPID software and its predictive ability for clinical functional outcomes are unclear.

    Therefore, we sought to compare the agreement of iStroke and RAPID on CTP imaging with ground truth on DWI and the association of infarct core volume (ICV) by softwares with clinical functional outcomes in patients with AIS who underwent endovascular treatment (EVT) with complete recanalisation.

    Methods

    Subjects

    We retrospectively included patients who were admitted to four teaching hospitals in China between January 2017 and September 2022. The inclusion criteria were as follows: (1) age ≥18 years old; (2) patient with AIS; (4) the patient underwent preprocedure CTP and postprocedure MRI examinations; (4) received EVT within 24 hours of symptom onset and achieved postprocedure complete recanalisation, which was defined as Modified Thrombolysis in Cerebral Infarction Score 316; (5) good imaging quality for evaluation of cerebral perfusion and ICV. Patients with incomplete or low-quality medical imaging were excluded. We collected patients’ demographics, clinical medical history (hypertension, diabetes mellitus, atrial fibrillation, etc), baseline National Institutes of Health Stroke Scale, treatment strategies (intravenous thrombolysis, EVT), perfusion imaging parameters and functional outcomes (Modified Rankin Scale (mRS) Score at 90 days). Excellent and favourable functional outcomes were defined as mRS ≤1 and mRS ≤2, respectively.

    Automated software on CTP

    Preprocedure CTP images of the same patients were processed using the image analysis software iStroke and RAPID independently. Infarct core and hypoperfusion volumes from the CTP data were automatically measured by the two softwares, respectively. Infarct core was defined as a CBF <30% of that in normal tissue, hypoperfusion volume was defined as Tmax >6 s. Large infarct core was defined as volume >70 mL.

    ICV and hypoperfusion data of RAPID software were retrospectively collected from local hospitals, while iStroke results were calculated by source imaging with Digital Imaging and Communications in Medicine (DICOMs) in core centre. iStroke uses digital phantoms17 of haemodynamics constructed from known perfusion parameter values. The architectural foundation of the model relies on recurrent neural networks, specifically implemented as a modified long short-term memory model, followed by a multilayer perceptron. First, a preprocessing pipeline, including motion correction, noise reduction, automatic AIF selection and brain tissue segmentation, is employed before inputting perfusion images into the deep learning model. Second, deep learning model generates the perfusion parameter maps undergoing threshold segmentation, resulting in the identification of regions corresponding to CBF <30% and Tmax >6 s. Third, during deployment, iStroke establishes a direct connection to imaging devices for image retrieval in the native DICOM format. Finally, the analysis results are sent back to the Picture Archiving and Communication Systems (PACS) for real-time review by clinical teams.

    Ground truth of infarct core

    We used the same method to manually trace the ground truth of infarct core on diffusion-weighted imaging (DWI). A 5-year experienced radiology researcher (YH) measured the infarct volume and was blinded to the patient’s medical information. The tracing maps were double-checked by the senior neurologist (YX).

    Statistical analyses

    The agreement of the two softwares on mismatch volumes was tested by Spearman’s rank correlation tests. Kappa tests were used for agreement on large infarct core or mismatch volume ≥15 mL. Bland-Altman plots were used to evaluate the agreement of ICV by each software on CTP and ground truth. Logistic regression models were used to determine the association of the ICV assessed by two softwares on CTP with the functional outcomes. Statistical analyses were performed by the SPSS V.22 software package (SPSS, Chicago, Illinois, USA).

    Results

    Subjects

    We included 326 patients with AIS with preprocedure CTP and complete recanalisation after thrombectomy. Among them, 228 underwent DWI examination and 40 of them had an infarct volume >70 mL. Clinical characteristics of the included patients are shown in table 1.

    View this table:
    Table 1

    Clinical characteristics of included patients across different subgroups

    Head-to-head comparison of perfusion volumes on automated softwares

    In all patients, iStroke was strongly correlated with RAPID on ICV (median 20.0 mL, IQR 5.0–35.0 mL vs median 6.0 mL, IQR 0.0–23.0 mL, respectively, ρ=0.68, p<0.001) and hypoperfusion volume (median 131.2 mL, IQR 67.9–183.5 mL vs median 135.5 mL, IQR 83.8–187.5 mL, respectively, ρ=0.66, p<0.001) (table 2). Boxplots illustrate the infarct core and hypoperfusion volumes of the two softwares (online supplemental figure 1).

    Supplemental material

    View this table:
    Table 2

    Agreement of automated softwares on infarct core volume (ICV) and hypoperfused volume for all patients

    Consistent with the main result, close correlation between iStroke and RAPID softwares was also found on hypoperfusion volume and ICV in anterior and posterior circulation stroke separately (online supplemental table 1).

    Supplemental material

    Substantial agreement was observed between these two softwares for large infarct core (kappa=0.73, p<0.001). While the mismatch volume ≥15 mL on iStroke had a good agreement with that on RAPID (kappa=0.35, p<0.001) (table 2). With regard to the mismatch ratio ≥1.8, 308 patients were consistent on both softwares. One patient was found to have a mismatch ratio <1.8 on both softwares, and other 17 cases had inconsistent findings between RAPID and iStroke. The kappa value was not suitable and the agreement rate was 94.79% (309/326).

    Association of ICV with functional outcomes

    Shift analysis by ordinal logistic regression revealed that ICV on iStroke and RAPID significantly correlated with 90-day mRS distribution after adjusting for age and sex (p=0.003 and p<0.001, respectively). The ICV on iStroke and RAPID were both significantly associated with an excellent to favourable functional outcomes with age and sex as covariates (table 3).

    View this table:
    Table 3

    The impact of ICV on functional outcomes

    Agreement of ICV on CTP softwares and DWI

    In patients with DWI examination, the ICV on both iStroke and RAPID softwares significantly correlated with the ground truth (ρ=0.43 and ρ=0.62, respectively) (table 4). Furthermore, strong correlation of ground truth with the two softwares was observed in patients with anterior large vessel occlusion (ρ of iStroke=0.43, ρ of RAPID=0.65, respectively), however, was not significant in patients with posterior large vessel occlusion (ρ of iStroke=0.18, ρ of RAPID=−0.016, respectively) (online supplemental table 2).

    View this table:
    Table 4

    Correlation of ICV on CTP softwares and DWI

    In patients with ground truth core volume >70 mL, the ICV on both iStroke and RAPID softwares was strongly correlated with the ground truth (ρ=0.51 and ρ=0.66, respectively) (table 4).

    With regards to individual agreement with ground truth, ICV on iStroke and RAPID was comparable to each other in patients with DWI (figure 1 and online supplemental figure 2) and in patients infarct volume >70 mL (figure 2).

    Supplemental material

    Figure 1
    Figure 1

    Agreement of infarct core volume between iStroke, RAPID and ground truth in patients with diffusion-weighted imaging. Volume difference between iStroke and ground truth in each patient was present as red dots (the mean volume difference is 14.0 mL, 95% CI −40.6 to 68.7 mL). Volume difference between RAPID and ground truth was present as blue dots (the mean volume difference is 21.8 mL, 95% CI −24.8 to 68.4 mL).

    Figure 2
    Figure 2

    Agreement of infarct core volume between iStroke, RAPID and ground truth in patients with diffusion-weighted imaging infarct volume >70 mL. Volume difference between iStroke and ground truth in each patient was present as red dots (the mean volume difference is 51.8 mL, 95% CI −9.2 to 112.8 mL). Volume difference between RAPID and ground truth was present as blue dots (the mean volume difference is 53.6 mL, 95% CI −7.1 to 114.2 mL).

    Discussion

    Our study revealed substantial agreement between iStroke and RAPID in measuring hypoperfusion and ICV on CTP. In patients with DWI, the ICV on both the iStroke and RAPID softwares was comparable in all patients and in large infarct core patients. The ICV on both softwares was significantly correlated with the patients’ functional outcomes at 90 days.

    Accurate estimation of perfusion volumes is critical to evaluate the amount of potentially salvageable tissue and provide timely treatment. In patients with AIS presenting within 6–24 hours of onset and with a large vessel occlusion in the anterior circulation, EVT is recommended when patients meet DAWN or DEFUSE 3 eligibility criteria.2 3 With a cut-off ICV of 70 mL per DEFUSE 3, overestimation of an ICV may inaccurately exclude patients from EVT, while underestimating of the ICV may be associated with poor clinical outcome following EVT.2 Therefore, the accuracy of the automatic image analysis software is crucial for case selection for EVT. In this study, we found that the iStroke had a substantial agreement with the RAPID in measuring the infarct core and hypoperfusion volumes from CTP data, both in anterior and posterior circulation large vessel occlusion. However, the agreement of mismatch volume ≥15 mL was only fair between the two softwares in all patients, probably due to a higher hypoperfusion volume and lower core volume on RAPID compared with those on iStroke. This finding was consistent with previous studies that showed RAPID tended to underestimate the ICV on CTP images compared with the ground truth4 18 19 and consequently overestimated the size of penumbra compared with iStroke.

    In patients with DWI examination, the ICV on both iStroke and RAPID softwares was significantly correlated with the ground truth. However, in the subgroup of posterior circulation arterial occlusion, this correlation was not significant for the two softwares. The contradictory results should be treated with caution. First, a limited sample size of only 29 patients who had a posterior circulation stroke was included. Second, ICV on the RAPID mismatch map tends to be characterised by significant low sensitivity and accuracy values, resulting in less adequate accuracy for detecting ischaemic lesions in posterior circulation regions.20 iStroke was not sensitive to posterior circulation infarcts either.

    Consistent with Shi et al study,21 we also found that the assessment of ICV measured by the iStroke was important in selecting patients for EVT and predicting their clinical outcome. Additionally, our study included the largest sample size of patients who underwent EVT with complete recanalisation,22 23 in order to control the possibility of infarct growth, to ensure the accuracy of ICV with the ground truth and to better illustrate the association of preprocedure ICV with postprocedure functional outcomes.

    As the most popular image processing platform, the RAPID software has been used for hypoperfusion-infarction mismatches in many large clinical trials and provided evidence for case selection for extended time-window thrombolysis and EVT. In addition, other commercially available CTP automated softwares including MIStar (Apollo Medical Imaging, Melbourne, Australia) and OLEA (Olea Medical Solutions, La Ciotat, France) are widely used in Australia and Europe. The iStroke has been installed in more than 100 clinical centres in China. The cross-comparison of these softwares is essential for future trials case selection and patients’ treatment strategies.

    Several limitations should be paid attention to. First, patients with AIS with large infarct core were limited, which may account for only a fair agreement between the ICV measured by the softwares and the ground truth, although it remained significant. Second, there was a time delay between the first CT scan and MRI (approximately 58 hours), during which the infarct volume might grow up. This change in volume may potentially lower the agreement of the assessment. Third, retrospective design of the study may limit the generalisability of the findings. Nonetheless, the included centres are large centres in China with common CTP modality parameters, and our study provided robust data on the accuracy of the CT perfusion volumes determined by iStroke. Further longitudinal studies with broader centres are needed.

    Conclusion

    The automatic CTP software iStroke is a reliable tool for assessing infarct core and mismatch volumes, making it clinically useful for selecting patients with AIS for acute reperfusion therapy in the extended time window.

    Data availability statement

    Data are available upon reasonable request.

    Ethics statements

    Patient consent for publication

    Ethics approval

    All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional research committee and the principles of the Declaration of Helsinki. The study was approved by the institutional review board and ethics committee at Beijing Tiantan Hospital without the requirement of patients’ written informed consents (KY2022-029-01).

    References

    1. ↵
      1. Xiong Y,
      2. Wakhloo AK,
      3. Fisher M
      . Advances in acute ischemic stroke therapy. Circ Res 2022;130:1230–51. doi:10.1161/CIRCRESAHA.121.319948
      OpenUrlCrossRef
    2. ↵
      1. Albers GW,
      2. Marks MP,
      3. Kemp S, et al
      . Thrombectomy for stroke at 6 to 16 hours with selection by perfusion imaging. N Engl J Med 2018;378:708–18. doi:10.1056/NEJMoa1713973
      OpenUrlCrossRefPubMed
    3. ↵
      1. Nogueira RG,
      2. Jadhav AP,
      3. Haussen DC, et al
      . Thrombectomy 6 to 24 hours after stroke with a mismatch between deficit and infarct. N Engl J Med 2018;378:11–21. doi:10.1056/NEJMoa1706442
      OpenUrlCrossRefPubMed
    4. ↵
      1. Austein F,
      2. Riedel C,
      3. Kerby T, et al
      . Comparison of perfusion CT software to predict the final infarct volume after Thrombectomy. Stroke 2016;47:2311–7. doi:10.1161/STROKEAHA.116.013147
      OpenUrlAbstract/FREE Full Text
    5. ↵
      1. Fieselmann A,
      2. Kowarschik M,
      3. Ganguly A, et al
      . Deconvolution-based CT and MR brain perfusion measurement: theoretical model Revisited and practical implementation details. Int J Biomed Imaging 2011:467563. doi:10.1155/2011/467563
    6. ↵
      1. Wirestam R,
      2. Andersson L,
      3. Ostergaard L, et al
      . Assessment of regional cerebral blood flow by dynamic susceptibility contrast MRI using different Deconvolution techniques. Magn Reson Med 2000;43:691–700. doi:10.1002/(sici)1522-2594(200005)43:5<691::aid-mrm11>3.0.co;2-b
      OpenUrlCrossRefPubMedWeb of Science
    7. ↵
      1. Ostergaard L,
      2. Weisskoff RM,
      3. Chesler DA, et al
      . High resolution measurement of cerebral blood flow using Intravascular Tracer bolus passages. Magn Reson Med 1996;36:715–25. doi:10.1002/mrm.1910360510
      OpenUrlCrossRefPubMedWeb of Science
    8. ↵
      1. Zanderigo F,
      2. Bertoldo A,
      3. Pillonetto G, et al
      . Nonlinear stochastic Regularization to characterize tissue residue function in bolus-tracking MRI: assessment and comparison with SVD, block-Circulant SVD, and Tikhonov. IEEE Trans Biomed Eng 2009;56:1287–97. doi:10.1109/TBME.2009.2013820
      OpenUrlPubMed
    9. ↵
      1. Wu O,
      2. Østergaard L,
      3. Weisskoff RM, et al
      . Tracer arrival timing-insensitive technique for estimating flow in MR perfusion-weighted imaging using singular value decomposition with a block-Circulant Deconvolution matrix. Magn Reson Med 2003;50:164–74. doi:10.1002/mrm.10522
      OpenUrlCrossRefPubMedWeb of Science
    10. ↵
      1. Kudo K,
      2. Sasaki M,
      3. Yamada K, et al
      . Differences in CT perfusion maps generated by different commercial software: quantitative analysis by using identical source data of acute stroke patients. Radiology 2010;254:200–9. doi:10.1148/radiol.254082000
      OpenUrlCrossRefPubMedWeb of Science
    11. ↵
      1. Andersen IK,
      2. Szymkowiak A,
      3. Rasmussen CE, et al
      . Perfusion Quantification using Gaussian process Deconvolution. Magn Reson Med 2002;48:351–61. doi:10.1002/mrm.10213
      OpenUrlCrossRefPubMed
    12. ↵
      1. Calamante F,
      2. Gadian DG,
      3. Connelly A
      . Quantification of bolus-tracking MRI: improved characterization of the tissue residue function using Tikhonov Regularization. Magn Reson Med 2003;50:1237–47. doi:10.1002/mrm.10643
      OpenUrlCrossRefPubMed
    13. ↵
      1. Mouridsen K,
      2. Friston K,
      3. Hjort N, et al
      . Bayesian estimation of cerebral perfusion using a physiological model of Microvasculature. Neuroimage 2006;33:570–9. doi:10.1016/j.neuroimage.2006.06.015
      OpenUrlCrossRefPubMedWeb of Science
    14. ↵
      1. Boutelier T,
      2. Kudo K,
      3. Pautot F, et al
      . Bayesian hemodynamic parameter estimation by bolus tracking perfusion weighted imaging. IEEE Trans Med Imaging 2012;31:1381–95. doi:10.1109/TMI.2012.2189890
      OpenUrlCrossRefPubMed
    15. ↵
      1. Xiong Y,
      2. Luo Y,
      3. Wang M, et al
      . Evaluation of diffusion-perfusion mismatch in acute ischemic stroke with a new automated perfusion-weighted imaging software: A retrospective study. Neurol Ther 2022;11:1777–88. doi:10.1007/s40120-022-00409-w
      OpenUrl
    16. ↵
      1. Yoo AJ,
      2. Simonsen CZ,
      3. Prabhakaran S, et al
      . Refining angiographic biomarkers of Revascularization: improving outcome prediction after intra-arterial therapy. Stroke 2013;44:2509–12. doi:10.1161/STROKEAHA.113.001990
      OpenUrlAbstract/FREE Full Text
    17. ↵
      1. Kudo K,
      2. Christensen S,
      3. Sasaki M, et al
      . Accuracy and reliability assessment of CT and MR perfusion analysis software using a Digital phantom. Radiology 2013;267:201–11. doi:10.1148/radiol.12112618
      OpenUrlCrossRefPubMed
    18. ↵
      1. Rava RA,
      2. Snyder KV,
      3. Mokin M, et al
      . Assessment of computed tomography perfusion software in predicting spatial location and volume of infarct in acute ischemic stroke patients: a comparison of sphere, Vitrea, and RAPID. J Neurointerv Surg 2021;13:130–5. doi:10.1136/neurintsurg-2020-015966
      OpenUrlAbstract/FREE Full Text
    19. ↵
      1. Wouters A,
      2. Robben D,
      3. Christensen S, et al
      . Prediction of stroke infarct growth rates by baseline perfusion imaging. Stroke 2022;53:569–77. doi:10.1161/STROKEAHA.121.034444
      OpenUrl
    20. ↵
      1. Capasso R,
      2. Vallone S,
      3. Serra N, et al
      . Qualitative versus automatic evaluation of CT perfusion parameters in acute posterior circulation ischaemic stroke. Neuroradiology 2021;63:317–30. doi:10.1007/s00234-020-02517-6
      OpenUrlPubMed
    21. ↵
      1. Shi Z,
      2. Li J,
      3. Zhao M, et al
      . Baseline cerebral ischemic core quantified by different automatic software and its predictive value for clinical outcome. Front Neurosci 2021;15:608799. doi:10.3389/fnins.2021.608799
    22. ↵
      1. Yedavalli V,
      2. Kihira S,
      3. Shahrouki P, et al
      . CTP-based estimated ischemic core: a comparative multicenter study between Olea and RAPID software. J Stroke Cerebrovasc Dis 2023;32:107297. doi:10.1016/j.jstrokecerebrovasdis.2023.107297
      OpenUrl
    23. ↵
      1. Yang W,
      2. Hoving JW,
      3. Koopman MS, et al
      . Agreement between estimated computed tomography perfusion ischemic core and follow-up infarct on diffusion-weighted imaging. Insights Imaging 2022;13:191. doi:10.1186/s13244-022-01334-0

    Supplementary material

    Footnotes

    • YX and YW contributed equally.

    • Contributors Study concept and design: YW and YX. Acquisition of data: PY, HW, ZX, FC, YH, LW, MH and NW. Analysis or interpretation of data: ZC, K-XY and WY. ZC and XF wrote the first draft of the manuscript. Critical revision of the manuscript for important intellectual content: DW, YW and YX. YW and YX, as guarantors, were responsible for the overall content. All authors reviewed the final version.

    • Funding This study is supported by Beijing Municipal Science & Technology Committee (Z211100003521019).

    • Competing interests None declared.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.