Artificial intelligence for decision support in acute stroke — current roles and potential


The identification and treatment of patients with stroke is becoming increasingly complex as more treatment options become available and new relationships between disease features and treatment response are continually discovered. Consequently, clinicians must constantly learn new skills (such as clinical evaluations or image interpretation), stay up to date with the literature and incorporate advances into everyday practice. The use of artificial intelligence (AI) to support clinical decision making could reduce inter-rater variation in routine clinical practice and facilitate the extraction of vital information that could improve identification of patients with stroke, prediction of treatment responses and patient outcomes. Such support systems would be ideal for centres that deal with few patients with stroke or for regional hubs, and could assist informed discussions with the patients and their families. Moreover, the use of AI for image processing and interpretation in stroke could provide any clinician with an imaging assessment equivalent to that of an expert. However, any AI-based decision support system should allow for expert clinician interaction to enable identification of errors (for example, in automated image processing). In this Review, we discuss the increasing importance of imaging in stroke management before exploring the potential and pitfalls of AI-assisted treatment decision support in acute stroke.

Key points

  • Imaging-based treatment guidance has been demonstrated as an effective approach in patients with a suspected stroke.

  • Clinical trials in which imaging is not used for patient selection are likely to include many patients with minor stroke or stroke mimics, making treatment effects difficult to detect.

  • Artificial intelligence (AI) and machine learning could provide image interpretation that equals or exceeds that of experts and could collate key features to assist clinicians with treatment decisions.

  • AI could be used to generate estimations of likely patient outcomes, which would not only be useful for assisting treatment decisions but also for informing family discussions.

  • AI-based decision assistance systems could be especially useful for centres without dedicated stroke specialists.

  • Any decision assistance tools must be validated and applied appropriately, and clear guidelines are needed to define how useful systems are in clinical practice.

Access options

Rent or Buy article

Get time limited or full article access on ReadCube.


All prices are NET prices.

Fig. 1: Perfusion imaging outputs from validated automated post-processing software.
Fig. 2: Treatment decision making by humans alone and in combination with artificial intelligence.
Fig. 3: Irreversible injury in ischaemic stroke.
Fig. 4: Identification of patients who are unlikely to benefit from therapy.
Fig. 5: Clinical implications of poor image acquisition.
Fig. 6: A poorly acquired acute CT perfusion scan.


  1. 1.

    Goyal, M. et al. Endovascular thrombectomy after large-vessel ischaemic stroke: a meta-analysis of individual patient data from five randomised trials. Lancet 387, 1723–1731 (2016). An excellent meta-analysis of thrombectomy trials that demonstrates a very large treatment effect.

    Article  Google Scholar 

  2. 2.

    Potter, C. A. et al. CT for treatment selection in acute ischemic stroke: a code stroke primer. Radiographics 39, 6 (2019).

    Google Scholar 

  3. 3.

    Sanelli, P. C. et al. Imaging and treatment of patients with acute stroke: an evidence-based review. Am. J. Neuroradiol. 35, 1045–1051 (2014).

    CAS  PubMed  Google Scholar 

  4. 4.

    Bacharach, R., Niazi, M. & Ermak, D. Pitfalls of CT perfusion imaging in acute ischemic stroke: a case series. Neurology 90 (Suppl. 15), P3.206 (2018).

    Google Scholar 

  5. 5.

    Adeoye, O. et al. Recommendations for the establishment of stroke systems of care: a 2019 update. A policy statement from the American Stroke Association. Stroke 50, e187–e210 (2019).

    PubMed  Google Scholar 

  6. 6.

    Heit, J. & Wintermark, M. Perfusion computed tomography for the evaluation of acute ischemic stroke: strengths and pitfalls. Stroke 47, 1153–1158 (2016).

    PubMed  Google Scholar 

  7. 7.

    Gao, J. et al. Visibility of CT early ischemic change is significantly associated with time from stroke onset to baseline scan beyond the first 3 hours of stroke onset. J. Stroke 19, 340–346 (2017).

    PubMed  PubMed Central  Google Scholar 

  8. 8.

    Fiebach, J. B. et al. CT and diffusion-weighted MR imaging in randomized order: diffusion-weighted imaging results in higher accuracy and lower interrater variability in the diagnosis of hyperacute ischemic stroke. Stroke 33, 2206–2210 (2002).

    CAS  PubMed  Google Scholar 

  9. 9.

    Logallo, N. et al. Tenecteplase versus alteplase for management of acute ischaemic stroke (NOR-TEST): a phase 3, randomised, open-label, blinded endpoint trial. Lancet Neurol. 16, 781–788 (2017).

    CAS  PubMed  Google Scholar 

  10. 10.

    IST-3 collaborative group. et al. The benefits and harms of intravenous thrombolysis with recombinant tissue plasminogen activator within 6 h of acute ischaemic stroke (the third international stroke trial [IST-3]): a randomised controlled trial. Lancet 379, 2352–2363 (2012).

    Google Scholar 

  11. 11.

    Bivard, A. et al. Too good to treat? Ischemic stroke patients with small computed tomography perfusion lesions may not benefit from thrombolysis. Ann. Neurol. 80, 286–293 (2016). A controversial article in which the benefit of thrombolysis for patients with a perfusion lesion <15 ml is questioned.

    CAS  PubMed  Google Scholar 

  12. 12.

    Nogueira, R. G. et al. Thrombectomy 6 to 24 hours after stroke with a mismatch between deficit and infarct. N. Engl. J. Med. 378, 11–21 (2018).

    PubMed  Google Scholar 

  13. 13.

    Albers, G. et al. Thrombectomy for stroke at 6 to 16 hours with selection by perfusion imaging. N. Engl. J. Med. 378, 11–21 (2018).

    Google Scholar 

  14. 14.

    Thomalla, G. et al. MRI-guided thrombolysis for stroke with unknown time of onset. N. Engl. J. Med. 379, 611–622 (2018). Together with Albers et al. (2018), these papers represent medical breakthroughs as they show that the treatment time window for stroke caused by large-vessel occlusion can be extended from 6 h to 24 h with the aid of imaging-based patient selection.

    PubMed  Google Scholar 

  15. 15.

    Ma, H. Extending thrombolysis to 9 hours and wake-up stroke using perfusion imaging selection. N. Engl. J. Med. 380, 1795–1803 (2019). A clinical trial of thrombolysis that shows that the treatment time window can be extended to 9 h or that the treatment can be used for patients with an unknown time of onset but have an imaging treatment target seen on CT perfusion.

    PubMed  Google Scholar 

  16. 16.

    Mokin, M. et al. Predictive value of RAPID assessed perfusion thresholds on final infarct volume in SWIFT PRIME (Solitaire with the intention for thrombectomy as primary endovascular treatment). Stroke 48, 932–938 (2017).

    PubMed  Google Scholar 

  17. 17.

    Hoving, J. W. et al. Volumetric and spatial accuracy of computed tomography perfusion estimated ischemic core volume in patients with acute ischemic stroke. Stroke 49, 2368–2375 (2018).

    PubMed  Google Scholar 

  18. 18.

    Campbell, B. C. et al. Cerebral blood flow is the optimal CT perfusion parameter for assessing infarct core. Stroke 42, 3435–3440 (2011).

    PubMed  Google Scholar 

  19. 19.

    Huang, X. et al. Alteplase versus tenecteplase for thrombolysis after ischaemic stroke (ATTEST): a phase 2, randomised, open-label, blinded endpoint study. Lancet Neurol. 14, 368–376 (2015).

    CAS  PubMed  Google Scholar 

  20. 20.

    Parsons, M. et al. A randomized trial of tenecteplase versus alteplase for acute ischemic stroke. N. Engl. J. Med. 366, 1099–1107 (2012).

    CAS  PubMed  Google Scholar 

  21. 21.

    Bivard, A. et al. Impact of computed tomography perfusion imaging on the response to tenecteplase in ischemic stroke: analysis of 2 randomized controlled trials. Circulation 135, 440–448 (2017).

    CAS  PubMed  Google Scholar 

  22. 22.

    Takahashi, N. et al. An automated detection method for the MCA dot sign of acute stroke in unenhanced CT. Radiol. Phys. Technol. 7, 79–88 (2014).

    PubMed  Google Scholar 

  23. 23.

    Bivard, A. et al. Perfusion computed tomography to assist decision making for stroke thrombolysis. Brain 138, 1919–1931 (2015).

    PubMed  PubMed Central  Google Scholar 

  24. 24.

    Hussein, H. M., Saleem, M. A. & Qureshi, A. I. Rates and predictors of futile recanalization in patients undergoing endovascular treatment in a multicenter clinical trial. Neuroradiology 60, 557–563 (2018).

    PubMed  Google Scholar 

  25. 25.

    Russell, S. J. & Norvig, P. Artificial Intelligence: a Modern Approach 3rd edn (Prentice Hall, 2009).

  26. 26.

    SAS. Machine learning: what it is and why it matters. (2020).

  27. 27.

    Domingos, P. A few useful things to know about machine learning. Commun. ACM. 55, 78–87 (2012).

    Google Scholar 

  28. 28.

    Brzdok, D., Krzywinski, M. & Altman, N. Machine learning: a primer. Nat. Methods. 14, 1119–1120 (2017).

    Google Scholar 

  29. 29.

    von Kummer, R. Ischemic stroke and tissue hypodensity on computed tomography. Stroke 30, 1974 (1999).

    Google Scholar 

  30. 30.

    Chen, Y. et al. Automated quantification of cerebral edema following hemispheric infarction: application of a machine-learning algorithm to evaluate CSF shifts on serial head CTs. Neuroimage Clin. 12, 673–680 (2016).

    PubMed  PubMed Central  Google Scholar 

  31. 31.

    Chen, L., Bentley, P. & Rueckert, D. Fully automatic acute ischemic lesion segmentation in DWI using convolutional neural networks. Neuroimage Clin. 15, 633–643 (2017).

    PubMed  PubMed Central  Google Scholar 

  32. 32.

    Meretoja, A., Keshtkaran, M., Tatlisumak, T., Donnan, G. A. & Churilov, L. Endovascular therapy for ischemic stroke: save a minute–save a week. Neurology 88, 2123–2127 (2017).

    PubMed  Google Scholar 

  33. 33.

    Meretoja, A. et al. Stroke thrombolysis: save a minute, save a day. Stroke 45, 1053–1058 (2014).

    PubMed  Google Scholar 

  34. 34.

    Pidd, M. Why modelling and model use matter. J. Oper. Res. Soc. 61, 14–24 (2010).

    Google Scholar 

  35. 35.

    Tan, J. C. et al. Systematic comparison of perfusion CT and CT-angiography in acute stroke patients. Ann. Neurol. 61, 533–543 (2007).

    PubMed  Google Scholar 

  36. 36.

    Maier, O., Schröder, C., Forkert, N. D., Martinetz, T. & Handels, H. Classifiers for ischemic stroke lesion segmentation: a comparison study. PLoS ONE 10, e0145118 (2015).

    PubMed  PubMed Central  Google Scholar 

  37. 37.

    Wilke, M., de Haan, B., Juenger, H. & Karnath, H. O. Manual, semi-automated, and automated delineation of chronic brain lesions: a comparison of methods. Neuroimage 56, 2038–2046 (2011).

    PubMed  Google Scholar 

  38. 38.

    Barber, P. A., Demchuk, A. M., Zhang, J. & Buchan, A. M. Validity and reliability of a quantitative computed tomography score in predicting outcome of hyperacute stroke before thrombolytic therapy. ASPECTS study group. Alberta stroke programme early CT score. Lancet 355, 1670–1674 (2000).

    CAS  PubMed  Google Scholar 

  39. 39.

    Kobkitsuksakul, C., Tritanon, O. & Suraratdecha, V. Interobserver agreement between senior radiology resident, neuroradiology fellow, and experienced neuroradiologist in the rating of Alberta stroke program early computed tomography score (ASPECTS). Diagn. Interv. Radiol. 24, 104–107 (2018).

    PubMed  PubMed Central  Google Scholar 

  40. 40.

    Nagel, S. 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 12, 615–622 (2017).

    PubMed  Google Scholar 

  41. 41.

    Austein, F. et al. Automated versus manual imaging assessment of early ischemic changes in acute stroke: comparison of two software packages and expert consensus. Eur. Radiol. 29, 6285–6292 (2019).

    PubMed  Google Scholar 

  42. 42.

    Castle, J. et al. Agreement regarding diagnosis of transient ischemic attack fairly low among stroke-trained neurologists. Stroke 41, 1367–1370 (2010).

    PubMed  Google Scholar 

  43. 43.

    Bivard, A. et al. Ischemic core thresholds change with time to reperfusion: a case control study. Ann. Neurol. 82, 995–1003 (2017).

    CAS  PubMed  PubMed Central  Google Scholar 

  44. 44.

    Sillanpaa, N. et al. The clot burden score, the Boston acute stroke imaging scale, the cerebral blood volume ASPECTS, and two novel imaging parameters in the prediction of clinical outcome of ischemic stroke patients receiving intravenous thrombolytic therapy. Neuroradiology 54, 663–672 (2012).

    PubMed  Google Scholar 

  45. 45.

    Bivard, A. et al. Validating a predictive model of acute advanced imaging biomarkers in ischemic stroke. Stroke 48, 645–650 (2017).

    PubMed  Google Scholar 

  46. 46.

    Kawano, H. et al. Perfusion CT in stroke thrombolysis patients: save a ml of penumbra, save a week of disability-free life. Brain 140, 684–691 (2017).

    PubMed  Google Scholar 

  47. 47.

    Liebeskind, D. Artificial intelligence in stroke care: deep learning or superficial insight? EBioMedicine 35, 14–15 (2018).

    PubMed  PubMed Central  Google Scholar 

  48. 48.

    Campbell, B. C. et al. Advanced imaging improves prediction of hemorrhage after stroke thrombolysis. Ann. Neurol. 73, 510–519 (2013).

    PubMed  PubMed Central  Google Scholar 

  49. 49.

    van Os, H. J. A. et al. Predicting outcome of endovascular treatment for acute ischemic stroke: potential value of machine learning algorithms. Front. Neurol. 9, 784 (2018).

    PubMed  PubMed Central  Google Scholar 

  50. 50.

    Bivard, A. et al. Global white matter hypoperfusion on CT predicts larger infarcts and hemorrhagic transformation after acute ischemia. CNS Neurosci. Ther. 22, 238–243 (2016).

    PubMed  PubMed Central  Google Scholar 

  51. 51.

    Campbell, B. C. 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 41, 82–88 (2010).

    PubMed  Google Scholar 

  52. 52.

    Yassi, N. et al. Prediction of poststroke hemorrhagic transformation using computed tomography perfusion. Stroke 44, 3039–3043 (2013).

    PubMed  Google Scholar 

  53. 53.

    Bentley, P. et al. Prediction of stroke thrombolysis outcome using CT brain machine learning. Neuroimage Clin. 4, 635–640 (2014).

    PubMed  PubMed Central  Google Scholar 

  54. 54.

    FDA. FDA permits marketing of clinical decision support software for alerting providers of a potential stroke in patients. (2018).

  55. 55.

    Wang, X. et al. ChestX-ray8: hospital-scale chest X-ray database and benchmarks on weakly-supervised classification and localization of common thorax diseases. IEEE CVPR (2017).

    Article  Google Scholar 

  56. 56.

    Titano, J. J. et al. Automated deep-neural-network surveillance of cranial images for acute neurologic events. Nat. Med. 24, 1337–1341 (2018).

    CAS  PubMed  Google Scholar 

  57. 57.

    Lobo, J. M., Jiménez-Valverde, A. & Real, R. AUC: a misleading measure of the performance of predictive distribution models. Glob. Ecol. Biogeogr. 17, 145–151 (2008). An important analytical piece that shows how to interpret statistical results.

    Google Scholar 

  58. 58.

    Saito, T. & Rehmsmeier, M. The precision–recall plot is more informative than the ROC plot when evaluating binary classifiers on imbalanced datasets. PLoS ONE 10, e0118432 (2015).

    PubMed  PubMed Central  Google Scholar 

Download references

Author information




All authors contributed to all aspects of the manuscript.

Corresponding author

Correspondence to Mark Parsons.

Ethics declarations

Competing interests

M.P. and A.B. have active research collaborations with Siemens Healthineers, Canon medical Systems and Apollo Medical Imaging. L.C. declares no competing interests.

Additional information

Peer review information

Nature Reviews Neurology thanks the anonymous reviewer(s) for their contribution to the peer review of this work.

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Bivard, A., Churilov, L. & Parsons, M. Artificial intelligence for decision support in acute stroke — current roles and potential. Nat Rev Neurol 16, 575–585 (2020).

Download citation


Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing