You are here

  • Home
  • Archive
  • Volume 16, Issue 12
  • Automated assessment of ischemic core on non-contrast computed tomography: a multicenter comparative analysis with CT perfusion

Article Text

Article menu
Ischemic stroke
Original research
Automated assessment of ischemic core on non-contrast computed tomography: a multicenter comparative analysis with CT perfusion
  1. Puja Shahrouki1,
  2. Shingo Kihira1,
  3. Elham Tavakkol1,
  4. Joe X Qiao1,
  5. Achala Vagal2,
  6. Pooja Khatri3,
  7. Mersedeh Bahr-Hosseini4,
  8. Geoffrey P Colby1,5,
  9. Reza Jahan1,
  10. Gary Duckwiler1,
  11. Viktor Szeder1,4,
  12. Luke Ledbetter1,
  13. Stephen Cai1,
  14. Banafsheh Salehi1,
  15. Amish H. Doshi6,
  16. Puneet Belani6,
  17. Johanna T Fifi7,
  18. Reade De Leacy7,
  19. J Mocco7,
  20. Jeffrey L Saver4,
  21. David S Liebeskind4,
  22. Kambiz Nael1
  1. 1 Department of Radiological Sciences, David Geffen School of Medicine, University of California Los Angeles, Los Angeles, California, USA
  2. 2 Department of Radiology, University of Cincinnati Medical Center, Cincinnati, Ohio, USA
  3. 3 Department of Neurology, University of Cincinnati, Cincinnati, Ohio, USA
  4. 4 Department of Neurology, David Geffen School of Medicine, University of California Los Angeles, Los Angeles, California, USA
  5. 5 Department of Neurosurgery, David Geffen School of Medicine, University of California Los Angeles, Los Angeles, California, USA
  6. 6 Department of Radiology, Icahn School of Medicine at Mount Sinai, New York, New York, USA
  7. 7 Department of Neurosurgery, Icahn School of Medicine at Mount Sinai, New York, New York, USA
  1. Correspondence to Dr Kambiz Nael; kambiznael{at}gmail.com

Abstract

Background Application of machine learning (ML) algorithms has shown promising results in estimating ischemic core volumes using non-contrast CT (NCCT).

Objective To assess the performance of the e-Stroke Suite software (Brainomix) in assessing ischemic core volumes on NCCT compared with CT perfusion (CTP) in patients with acute ischemic stroke.

Methods In this retrospective multicenter study, patients with anterior circulation large vessel occlusions who underwent pretreatment NCCT and CTP, successful reperfusion (modified Thrombolysis in Cerbral Infarction ≥2b), and post-treatment MRI, were included from three stroke centers. Automated calculation of ischemic core volumes was obtained on NCCT scans using ML algorithm deployed by e-Stroke Suite and from CTP using Olea software (Olea Medical). Comparative analysis was performed between estimated core volumes on NCCT and CTP and against MRI calculated final infarct volume (FIV).

Results A total of 111 patients were included. Estimated ischemic core volumes (mean±SD, mL) were 20.4±19.0 on NCCT and 19.9±18.6 on CTP, not significantly different (P=0.82). There was moderate (r=0.40) and significant (P<0.001) correlation between estimated core on NCCT and CTP. The mean difference between FIV and estimated core volume on NCCT and CTP was 29.9±34.6 mL and 29.6±35.0 mL, respectively (P=0.94). Correlations between FIV and estimated core volume were similar for NCCT (r=0.30, P=0.001) and CTP (r=0.36, P<0.001).

Conclusions Results show that ML-based estimated ischemic core volumes on NCCT are comparable to those obtained from concurrent CTP in magnitude and in degree of correlation with MR-assessed FIV.

  • stroke
  • CT perfusion
  • CT

Data availability statement

Data are available upon reasonable request.

https://doi.org/10.1136/jnis-2023-020954

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

    • Recent application of artificial intelligence has shown promising results in estimation of ischemic core volume from non-contrast CT (NCCT).

    WHAT THIS STUDY ADDS

    • Machine learning-based estimated ischemic core volumes on NCCT available commercially, performs favorably in comparison with concurrent CT perfusion (CTP) in a multicenter study design.

    HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

    • Volumetric analysis of ischemic core on NCCT similar to those obtained from CTP can increase access for treatment decision-making and provide a setting for assessment of infarct growth while enjoying the simplicity of NCCT.

    Introduction

    Neuroimaging plays a significant role in the management of patients with acute ischemic stroke.1 Among neuroimaging variables, the estimated ischemic core volume on baseline imaging has been widely used for treatment decision-making and prognostication purposes. The modern endovascular treatment (EVT) trials have used a modest volume of established ischemic core (50–70 mL) as a treatment eligibility criteria.1–3 Although treatment exclusion based on having a large core has been debated recently with the results of large core trials,4–6 ischemic core volume remains as a major eligibility criteria on the American Heart Association/American Stroke Association guidelines1 until future guidelines and expert consensus to determine the relevance of ischemic core estimation in treating patients with acute ischemic stroke. Another value of ischemic core assessment is in determination of infarct growth, which has important prognostic implications and also has been a major target of neuroprotective trials.7 To properly design a trial and assess the value of neuroprotection, infarct growth needs to be accurately quantified, for which accurate estimation of ischemic core is critical.

    Ischemic core can be assessed by non-contrast CT (NCCT), CT perfusion (CTP), or MRI. During the early time window of ischemia, MRI diffusion-weighted imaging (DWI) can depict cytotoxic edema, and CTP can identify severe hemodynamic changes within the ischemic bed. Since the differences between normal brain and tissue undergoing severe ischemia are significant on DWI and CTP, both modalities can provide reasonable estimation of ischemic core volume using a variety of threshold-based methodologies. The most common thresholds for quantitative core estimation are apparent diffusion coefficient values of <600–620 x 10-6 mm2/s for MRI8 and relative cerebral blood flow <30%9 for CTP, which have successfully been used for treatment selection in randomized clinical trials.2 3

    On NCCT, tissue experiencing severe ischemia will appear hypodense due to ionic edema.10 However, assessment of ischemic core on NCCT has been traditionally subjective or semiquantitative using Alberta Stroke Program Early CT Score (ASPECTS). The density changes between ischemic and normal brain can be subtle, especially when there are overlapping density values between ischemic core and other pathologies, such as leukoaraiosis. As a result, volumetric assessment of ischemic core on NCCT has not been tried until recent years.

    Recent advances in machine learning (ML) applications have shown promise in obtaining volumetric core assessment from NCCT, enabling ischemic core volume measurement on routinely obtained NCCT, a task which has traditionally been reserved for more advanced imaging techniques such as CTP and MRI.11 12 Among ML-based techniques, e-Stroke Suite software (Brainomix, Oxford, UK) is now commercially available and has shown promising results in calculation of the ischemic core volumes from NCCT.13–17

    In this study, we hypothesized that estimated ischemic core volumes on NCCT obtained via ML-based e-Stroke Suite will be comparable to those obtained from CTP. To broaden the generalizability of our results, we designed a multicenter retrospective study with the following aims: (1) to compare the pretreatment ischemic core volumes obtained automatically from concurrent NCCT with CTP; (2) to perform a volumetric comparison of estimated ischemic core volumes between NCCT and CTP with MRI-final infarct volume (FIV) in successfully reperfused patients.

    Methods

    Study design

    This retrospective, multicenter study from three comprehensive stroke centers across the USA was carried out under an approved protocol by the institutional review boards of all three centers. From July 2015 to December 2020, patients with anterior circulation large vessel occlusions from three comprehensive stroke centers across the USA were included if: (1) they had pretreatment NCCT and CTP, (2) had undergone successful EVT (Thrombolysis in cerebral infarction scale (TICI) ≥2b), and (3) had post-treatment MRI.

    Patients with TICI<2b, unavailable or non-diagnostic CT and MR imaging, or studies with postprocessing software errors were excluded. Demographic and baseline clinical data, including age, sex, stroke-onset time, time of CT and MRI, baseline National Institutes of Health Stroke Scale (NIHSS) scores, and final reperfusion status defined by the TICI scale,18 were collected for each patient.

    NCCT and CTP imaging parameters for each site are detailed in online supplemental materials.

    Supplemental material

    Image analysis

    Ischemic core volumes from CTP were automatically calculated using Olea automated platform AP 1.5 (Olea Medical Solutions, La Ciotat, France), which uses a combination of relative cerebral blood flow <25% and differential time-to-peak >5 s as thresholds to calculate ischemic core volumes.

    Ischemic core volumes from NCCT were automatically calculated using e-Stroke Suite (Brainomix, Oxford, UK). Details of the algorithm have been previously reported.13 14 In short, the axial 5 mm NCCT digital imaging and communication in medicine (DICOM) images were used as the input, followed by image resampling and registration to standardize the input resolution and remove any tilt and rotation. The algorithm then uses a trained ML classifier to identify acute ischemia and derive a voxel-wise probability map, which is then segmented to provide volume of ischemic core.

    Assessment of ischemic core volume from NCCT on e-Stroke Suite and from CTP on Olea software was completely automated with no adjustment or human interface.

    Final infarct volumes from MRI-DWI were calculated on Olea Sphere, SP23 (Olea Medical Solutions, La Ciotat, France). A voxel-based signal intensity method subsuming the entire region of DWI hyperintensity was used to calculate volume of ischemic core. Estimated ischemic core volumes obtained from NCCT and CTP were then compared against MRI FIV.

    Statistical analysis

    Continuous data are expressed as means with SD or median with IQR, and categorical data are presented as absolute values and relative frequencies (percentages). Core volumes were assessed using Bland-Altman plots to evaluate potential differences between imaging techniques (NCCT, CTP, and MRI) and to assess the degree of agreement. Correlations between the core volumes obtained from the different imaging techniques were then tested using the Pearson’s test. All statistical analyses were conducted with the SPSS (version 28.0; SPSS, Chicago, Illinois, USA).

    Results

    Clinical characteristics of the patient population

    Our final cohort consisted of 111 patients. A total of 27 patients were excluded owing to poor image quality of CTP scans due to poor contrast bolus or motion (n=14), poor NCCT image quality due to motion or incomplete head coverage (n=4), poor image quality of MRI-DWI (n=5), failure of automated processing due to corrupted imaging files (n=4). The baseline demographic and clinical data are listed in table 1. The time (median, IQR) from last known well (LKW) to CTP was 1.6 (1.0–5.7) hours and the time (median, IQR) from CTP to reperfusion was 1.6 (1.2–2.2) hours. The time (median, IQR) from EVT to MRI was 24 (17.9–33.0) hours.

    View this table:
    Table 1

    Demographics, clinical and imaging characteristics

    Comparative analysis of ischemic core volume on NCCT vs CTP

    The estimated ischemic core volume (mean±SD) was 20.4±19.0 mL on NCCT and 19.9±18.6 mL on CTP. There was no significant difference (P=0.824) between NCCT- and CTP-estimated core volumes. There was moderate (r=0.40) but statistically significant (P<0.001) correlation between estimated core on NCCT and CTP. Figure 1 shows Bland-Altman and correlation plots between estimated ischemic core volume on NCCT and CTP. Subgroup analysis in patients who presented early (≤6 hours from LKW to imaging) compared with those who presented in the late window (>6 hours from LKW to imaging), showed no significant difference between estimated core volume on NCCT and CTP. The mean difference of estimated core between CTP and NCCT in patients who presented within 6 hours was 0±18.5 mL (P=0.979), and 1.8±27.3 mL (P=0.740) in those who presented beyond 6 hours (figure 2).

    Figure 1
    Figure 1

    (A) Bland-Altman plots showing mean differences and agreement for limits between NCCT (Brainomix) core estimation and CTP (Olea) core estimation. (B) Scatterplot showing the Pearson correlation between NCCT and CTP estimated core volumes. CTP, CT perfusion; NCCT, non-contrast CT.

    Figure 2
    Figure 2

    Bland-Altman plots showing mean differences and agreement for limits and scatterplot with Pearson correlation between NCCT (Brainomix) and CTP (Olea) core estimation in the early (A–B) and late (C–D) time windows. CTP, CT perfusion; NCCT, non-contrast CT.

    Comparative analysis of estimated ischemic core and final infarct volume

    Final infarction volume (mean±SD) obtained from post-treatment MRI-DWI was 39.3±61.0 mL. Bland-Altman and correlation plots for demonstration of the degree of agreement between FIV and estimated core volumes on NCCT and CTP are shown in figure 3. The mean difference between FIV and estimated core volume on NCCT and CTP was 29.9±34.6 mL and 29.6±35.0 mL, respectively (P=0.942). There was modest but statistically significant correlation between FIV and estimated core volume on NCCT (r=0.30, P=0.001) and CTP (r=0.36, P<0.001).

    Figure 3
    Figure 3

    Bland-Altman plot showing mean difference and agreement for limits between NCCT (Brainomix) and MR final infarct volume (FIV) (A) and CTP (Olea) core estimation and FIV (B). CTP, CT perfusion; NCCT, non-contrast CT.

    Figure 4 provides examples of ischemic core volumes segmented on NCCT and CTP in comparison with DWI-MRI.

    Figure 4
    Figure 4

    Examples of NCCT and CTP-based ischemic core estimation in two patients. (A) Patient in their 50 s with NIHSS score of 19 and ASPECTS of 8 with right internal carotid artery large vessel occlusion. NCCT-based calculated ischemic core volume was 24 mL, CTP-based volume was 29 mL, and final infarct volume (FIV) on MR was 30 mL. (B) Patient in their 70 s with NIHSS score of 15 and ASPECTS of 9 with left M2 territory large vessel occlusion. NCCT-based calculated ischemic core volume was 14 mL, CTP-based volume was 10 mL, and FIV was 11 mL. ASPECTS, Alberta Stroke Program Early CT Score; CTP, CT perfusion; NCCT, non-contrast CT; NIHSS, National Institutes of Health Stroke Scale.

    Discussion

    Artificial intelligence has been used to provide automated ASPECTS from NCCT with some success, aiming to reduce the potential inter-rater variability, with recent studies demonstrating non-inferiority of automated ASPECTS in comparison with neuroradiologist consensus scoring.19 20 Although ASPECTS are extremely useful for treatment decision-making in patients presenting within 6 hours of stroke-symptoms,1 obtaining volumetric assessment of ischemic core provide additional values similar to those obtain from CTP or MRI that can be used for EVT eligibility in patients with acute ischemic stroke presenting in the late window. Another important implication for volumetric ischemic core evaluation is to assess infarct growth, which is an important target of neuroprotective trials.7 For these reasons, the ability to obtain ischemic core volume from NCCT has been the target of recent investigations using artificial intelligence and ML-based techniques.12–16 21 22

    In our study, using commercially available software (e-Stroke, Brainomix) that uses a ML-based algorithm, we obtained volumetric ischemic core volumes from NCCT, which were not significantly different from those obtained from concurrent CTP and with comparable agreement with FIV derived from MRI-DWI. We included patients from three different stroke centers with variable CT acquisition parameters (see online supplemental material) to broaden the generalizability of our results. Our subgroup analysis accounting for the time-to stroke symptoms (LKW to initial imaging ≤6 hours vs >6 hours) found no significant difference in estimated core volumes between NCCT and CTP, indicating that the information gained could be used for treatment decision-making in a similar fashion. A recent study by Bouslama et al revealed similar moderate correlations between NCCT and CTP core estimation across early and late time windows, although NCCT core volumes were overall estimated as higher than CTP-estimated core volumes.13 In our study, however, the NCCT and CTP core volumes were closely similar, with very low mean differences (0±18.5 mL and 1.8±27.3 mL, for early and late windows, respectively).

    When comparing the Bland-Altman plots obtained in our study, an interesting observation can be made. In a total of seven patients there was a considerable discordance (defined as±2 SD) in estimated core between NCCT and CTP with an absolute mean difference of 52.2 mL (range 26.2–77.3 mL) in comparison with MRI-FIV. Among these patients, the discordance was related to erroneous estimation of core on NCCT in four patients and on CTP in the other three patients. Among four patients with erroneous estimation of NCCT core, two patients had significantly underestimated core probably due to insensitivity of the NCCT to early ischemic changes as they presented within 2 hours of LKW. In the other two patients, significant overestimation of NCCT core occurred due to white matter changes and streak artifacts. Among the three patients with erroneous estimation of core on CTP, one may be related to a known phenomenon of ghost core23 in the hyperacute setting as this patient presented within 1 hour from the stroke onset. The cause of overestimation of core on the other two CTP studies was not clear.

    Our results demonstrate the ability of a ML-based algorithm to estimate ischemic core volumes from NCCT with similar performance to CTP and may be an attractive alternative in centers without readily access to advanced imaging modalities.24 However, as Guberina et al highlight, great care needs to be taken to ensure automated assessments are supervised by an experienced radiologist, as pre-existing cerebral changes, such as leukoaraiosis, old infarct, and aneurysm coils, might not be effectively distinguished from acute ischemic core.25

    Our study has several limitations. First, it was a retrospective study, which may introduce inherent selection bias. With a prospective study, recruitment of patients with larger infarct volume would improve distribution of the infarct volumes and accuracy of our results. Second, NCCT and especially CTP protocols and acquisitions may vary across different sites and scanners, which might introduce confounding variables. However, variability in scanner brands and acquisition methods probably reflect daily clinical practice, supporting the applicability of our findings. Furthermore, this study only validates NCCT ML algorithms for e-Stroke Suite against estimated core volume obtained from Olea software. Further studies are required to determine if NCCT can be similarly used across different vendors before incorporation into management pathways. Lastly, post-treatment MR was used as the gold standard for FIV, which inherently limits the accuracy of this study as true infarct volume can change owing to the potential for continued infarct growth between CT (NCCT or CTP) and post-treatment MR despite successful reperfusion. We tried to minimize this factor by including only patients with successful reperfusion (TICI≥2b) and by showing extensive comparisons of NCCT scans with CTP scans which were acquired at the same time.

    In conclusion, the results show that estimated ischemic core volumes obtained automatically by e-Stroke software on NCCT are comparable to those obtained by concurrent CTP and correlate with post-treatment MR final infarct volume.

    Data availability statement

    Data are available upon reasonable request.

    Ethics statements

    Patient consent for publication

    Ethics approval

    Institutional review board of the Mount Sinai School of Medicine, HS#: 18-00789.

    Ethics approval

    This study involves human participants and was approved by the Institutional Review Board of the Mount Sinai School of Medicine, HS#: 18-00789.

    References

    1. ↵
      2. Powers WJ ,
      3. Rabinstein AA ,
      4. Ackerson T , et al
      . Guidelines for the early management of patients with acute ischemic stroke: 2019 update to the 2018 guidelines for the early management of acute ischemic stroke: a guideline for healthcare professionals from the American Heart Association/American Stroke Association. Stroke 2019;50:e344–418. doi:10.1161/STR.0000000000000211
      OpenUrlCrossRefPubMed
    2. ↵
      2. Albers GW ,
      3. Marks MP ,
      4. 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. ↵
      2. Nogueira RG ,
      3. Jadhav AP ,
      4. 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. ↵
      2. Yoshimura S ,
      3. Sakai N ,
      4. Yamagami H , et al
      . Endovascular therapy for acute stroke with a large ischemic region. N Engl J Med 2022;386:1303–13. doi:10.1056/NEJMoa2118191
      OpenUrlCrossRefPubMed
    5. ↵
      2. Sarraj A ,
      3. Pujara DK ,
      4. Campbell BC
      . Endovascular thrombectomy for acute large ischemic strokes. reply. N Engl J Med 2023;389:89–90. doi:10.1056/NEJMc2305915
      OpenUrl
    6. ↵
      2. Huo X ,
      3. Ma G ,
      4. Tong X , et al
      . Trial of endovascular therapy for acute ischemic stroke with large infarct. N Engl J Med 2023;388:1272–83. doi:10.1056/NEJMoa2213379
      OpenUrlPubMed
    7. ↵
      2. Goyal M ,
      3. McTaggart R ,
      4. Ospel JM , et al
      . How can imaging in acute ischemic stroke help us to understand tissue fate in the era of endovascular treatment and cerebroprotection Neuroradiology 2022;64:1697–707. doi:10.1007/s00234-022-03001-z
      OpenUrlCrossRefPubMed
    8. ↵
      2. Purushotham A ,
      3. Campbell BCV ,
      4. Straka M , et al
      . Apparent diffusion coefficient threshold for delineation of ischemic core. Int J Stroke 2015;10:348–53. doi:10.1111/ijs.12068
      OpenUrlCrossRefPubMed
    9. ↵
      2. Campbell BCV ,
      3. Christensen S ,
      4. Levi CR , et al
      . Cerebral blood flow is the optimal CT perfusion parameter for assessing infarct core. Stroke 2011;42:3435–40. doi:10.1161/STROKEAHA.111.618355
      OpenUrlAbstract/FREE Full Text
    10. ↵
      2. von Kummer R ,
      3. Dzialowski I
      . Imaging of cerebral ischemic edema and neuronal death. Neuroradiology 2017;59:545–53. doi:10.1007/s00234-017-1847-6
      OpenUrlPubMed
    11. ↵
      2. Cauley KA ,
      3. Mongelluzzo GJ ,
      4. Fielden SW
      . Automated estimation of acute infarct volume from noncontrast head CT using image intensity inhomogeneity correction. Int J Biomed Imaging 2019;2019:1720270. doi:10.1155/2019/1720270
    12. ↵
      2. Qiu W ,
      3. Kuang H ,
      4. Teleg E , et al
      . Machine learning for detecting early infarction in acute stroke with non-contrast-enhanced CT. Radiology 2020;294:638–44. doi:10.1148/radiol.2020191193
      OpenUrlPubMed
    13. ↵
      2. Bouslama M ,
      3. Ravindran K ,
      4. Harston G , et al
      . Noncontrast computed tomography E-stroke infarct volume is similar to RAPID computed tomography perfusion in estimating postreperfusion infarct volumes. Stroke 2021;52:634–41. doi:10.1161/STROKEAHA.120.031651
      OpenUrlCrossRefPubMed
    14. ↵
      2. Chen IE ,
      3. Tsui B ,
      4. Zhang H , et al
      . Automated estimation of ischemic core volume on noncontrast-enhanced CT via machine learning. Interv Neuroradiol 2022;2022:15910199221145487. doi:10.1177/15910199221145487
      OpenUrl
    15. ↵
      2. Jabal MS ,
      3. Joly O ,
      4. Kallmes D , et al
      . Interpretable machine learning modeling for ischemic stroke outcome prediction. Front Neurol 2022;13:884693. doi:10.3389/fneur.2022.884693
    16. ↵
      2. Suomalainen OP ,
      3. Elseoud AA ,
      4. Martinez-Majander N , et al
      . Comparison of automated infarct core volume measures between non-contrast computed tomography and perfusion imaging in acute stroke code patients evaluated for potential endovascular treatment. J Neurol Sci 2021;426:117483. doi:10.1016/j.jns.2021.117483
      OpenUrlPubMed
    17. ↵
      2. Nagel S ,
      3. Joly O ,
      4. Pfaff J , et al
      . e-ASPECTS derived acute ischemic volumes on non-contrast-enhanced computed tomography images. Int J Stroke 2020;15:995–1001. doi:10.1177/1747493019879661
      OpenUrlPubMed
    18. ↵
      2. Tomsick T ,
      3. Broderick J ,
      4. Carrozella J , et al
      . Revascularization results in the interventional management of stroke II trial. AJNR Am J Neuroradiol 2008;29:582–7. doi:10.3174/ajnr.A0843
      OpenUrlCrossRefPubMedWeb of Science
    19. ↵
      2. Nagel S ,
      3. Sinha D ,
      4. Day D , et al
      . e-ASPECTS software is non-inferior to neuroradiologists in applying the ASPECT score to computed tomography scans of acute ischemic stroke patients. Int J Stroke 2017;12:615–22. doi:10.1177/1747493016681020
      OpenUrlCrossRefPubMed
    20. ↵
      2. Sundaram VK ,
      3. Goldstein J ,
      4. Wheelwright D , et al
      . Automated ASPECTS in acute ischemic stroke: a comparative analysis with CT perfusion. AJNR Am J Neuroradiol 2019;40:2033–8. doi:10.3174/ajnr.A6303
      OpenUrlAbstract/FREE Full Text
    21. ↵
      2. Nishi H ,
      3. tsuji hirofumi ,
      4. Ishii A , et al
      . Abstract Tp98: deep learning-based automatic ischemic core estimation based on non-contrast-enhanced computed tomography. Stroke 2023;54. doi:10.1161/str.54.suppl_1.TP98
    22. ↵
      2. Wang C ,
      3. Shi Z ,
      4. Yang M , et al
      . Deep learning-based identification of acute ischemic core and deficit from non-contrast CT and CTA. J Cereb Blood Flow Metab 2021;41:3028–38. doi:10.1177/0271678X211023660
      OpenUrlPubMed
    23. ↵
      2. Boned S ,
      3. Padroni M ,
      4. Rubiera M , et al
      . Admission CT perfusion may overestimate initial infarct core: the ghost infarct core concept. J Neurointerv Surg 2017;9:66–9. doi:10.1136/neurintsurg-2016-012494
      OpenUrlAbstract/FREE Full Text
    24. ↵
      2. Pfaff J ,
      3. Herweh C ,
      4. Schieber S , et al
      . e-ASPECTS correlates with and is predictive of outcome after mechanical thrombectomy. AJNR Am J Neuroradiol 2017;38:1594–9. doi:10.3174/ajnr.A5236
      OpenUrlAbstract/FREE Full Text
    25. ↵
      2. Guberina N ,
      3. Dietrich U ,
      4. Radbruch A , et al
      . Detection of early infarction signs with machine learning-based diagnosis by means of the Alberta Stroke Program Early CT score (ASPECTS) in the clinical routine. Neuroradiology 2018;60:889–901. doi:10.1007/s00234-018-2066-5
      OpenUrlCrossRefPubMed

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Footnotes

    • X @rdeleacymd, @kambiznael

    • Presented at The results of this paper were presented at the ASNR 61st Annual Meeting, Chicago, Ilinois, USA. Abstract#473. April 29-May 3, 2023.

    • Contributors Study design: AV, JM, KN. Data collection: PS, SK, ET, JXQ, AV, PB, AD, JTF. Data analysis and interpretation: PS, PK, MB-H, GPC, RJ, GD, VS, LL, SC, BS, JTF, RDL, JM, JLS, DSL, KN. Clinical input: PK, MB-H, GPC, JLS, DSL, KN. Writing manuscript: PS, KN. Revising the manuscript for important intellectual content: All authors. Approved the final version to be published: All authors. Guarantor: KN.

    • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

    • Competing interests GPC: Stryker Neurovascular, MicroVention, Medtronic, Rapid Medical; JLS: Medtronic Cerenovus, Boehringer Ingelheim, NeuroVasc, Rapid Medical; DSL: Cerenovus, Genentech, Medtronic, Stryker, Olea; KN: Olea Medical, Brainomix; Achala Vagal: Viz AI.

    • 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.