Artificial Intelligence in “Code Stroke”—A Paradigm Shift: Do Radiologists Need to Change Their Practice?
2022; Radiological Society of North America; Volume: 4; Issue: 2 Linguagem: Inglês
10.1148/ryai.210204
ISSN2638-6100
Autores Tópico(s)Venous Thromboembolism Diagnosis and Management
ResumoHomeRadiology: Artificial IntelligenceVol. 4, No. 2 PreviousNext EditorialFree AccessArtificial Intelligence in "Code Stroke"—A Paradigm Shift: Do Radiologists Need to Change Their Practice?Achala Vagal , Luca SabaAchala Vagal , Luca SabaAuthor AffiliationsFrom the Department of Radiology, University of Cincinnati Medical Center, 234 Goodman St, Cincinnati, OH 45267-0525 (A.V.); and Department of Diagnostic Imaging and Radiology, University of Cagliari, Cagliari, Italy (L.S.).Address correspondence to A.V. (e-mail: [email protected]).Achala Vagal Luca SabaPublished Online:Jan 19 2022https://doi.org/10.1148/ryai.210204MoreSectionsPDF ToolsImage ViewerAdd to favoritesCiteTrack CitationsPermissionsReprints ShareShare onFacebookTwitterLinked In It is not the strongest of the species that survives, nor the most intelligent. It is the one that is most adaptable to change.– Charles DarwinIschemic stroke is one of the leading causes of mortality and severe disability worldwide (1). "Code stroke" is a time-sensitive and high-stakes clinical scenario alert for acute stroke that requires a rapid team approach to facilitate hyperacute evaluation and management of patients. For each acute stroke case not adequately interpreted, categorized, and treated, there is a high risk of mortality and disability.Management of acute ischemic stroke is evolving rapidly due to highly efficacious endovascular therapy (2–8). Regardless of the imaging modality (CT vs MRI) or type of hospital (tertiary "hub" vs outlying "spoke"), acute stroke management has one unifying need: to treat as quickly as possible. Optimal stroke treatment is a highly time-dependent process (9). Every small reduction in time to treatment helps bring lifetime benefits: every minute saved leads to 4 days of disability-free life (10,11). In this scenario, any technology is desirable that improves the diagnosis of stroke and rapidly informs treatment decisions.Artificial intelligence (AI) is rising as a leading component in stroke imaging and is expected to further change acute stroke care (12). There are three main use cases where AI offers potential benefits:Detection of abnormality to improve the performance of the human reader and increase the accuracy, sensitivity, and specificity of imaging analysis. AI can identify morphologic, mathematical, and geometric characteristics that are hard to detect with the human eye.Workflow efficiency to reduce the time to detect abnormalities and generate reports. AI can recognize the most severe conditions to expedite care and reduce treatment times.Therapeutic decision support to make triage and treatment decisions based on clinical and imaging parameters.There are three primary imaging questions in acute ischemic stroke that need to be answered quickly: extent of acute ischemia, presence of large vessel occlusion (LVO), and extent of brain tissue at risk. The Alberta Stroke Program Early CT Score (ASPECTS) provides a prognostic approach for extent of ischemia, where noncontrast head CT is scored by dividing the middle cerebral artery territory into 10 regions of interest (13). LVO accounts for one-third of acute ischemic stroke and can be assessed quickly with CT arteriography or MR arteriography. Perfusion imaging is another powerful tool that can measure ischemic and at-risk brain tissue. The constellation of imaging findings helps physicians make time-sensitive triage and treatment decisions, such as intravenous thrombolysis, endovascular treatment, patient transfer, or a decision not to treat.There are several practical limitations of using imaging tools for acute stroke. First, analysis of imaging information requires time. Second, not all radiologists are trained to analyze and interpret these techniques. Third, there is significantly limited availability of radiologists in community-based and rural hospitals that receive patients with acute stroke.To address these limitations, AI solutions have been built that can automate ASPECTS (14), LVO detection (15), and perfusion analysis. These AI solutions are being used currently in clinical practice with multiple Food and Drug Administration–approved and Conformité Européenne mark–certified commercially available platforms and are no longer a mere academic exercise (16,17). The AI application in stroke care is now integrated into clinical care with level 1 evidence in the American Heart Association/American Stroke Association guidelines (18). Additionally, in September 2020, the U.S. Centers for Medicare & Medicaid Services approved the first reimbursement for such AI-augmented medical care (19).AI-powered software platforms have streamlined stroke care workflows, reduced treatment times, and improved outcomes (12,17,20–22). Moreover, the effect of differing levels of imaging expertise is mitigated by AI models that offer the analysis automatically, which is even more critical for community-based and rural stroke centers that may not have the necessary 24/7 availability of expert radiology interpretation. Delays to treatment are particularly prevalent when patients require a transfer from hospitals that lack endovascular therapy capability onsite. The AI tools can help expedite this workflow and be cost-effective. A recent cost-effectiveness study of automated LVO detection–applied base case scenario (6% missed diagnosis, $40 per AI analysis, and 50% reduction of missed LVOs by AI) showed substantial cost savings and increased quality-adjusted life-years for patients with acute ischemic stroke over their projected lifetime (23).AI has drastically changed the acute stroke workflow. The AI technology runs in parallel with hospital picture archiving and communication systems (PACS) in which imaging data are transferred simultaneously to the AI cloud and to PACS. The imaging results are available within minutes on mobile devices or desktop workstations. Current cloud-based AI applications use nondiagnostic image viewers that provide easy access to CT or MRI on a smartphone (24–26). An alert about a new case notification is sent via mobile devices with rapid sharing of Health Insurance Portability and Accountability Act–compliant Digital Imaging and Communications in Medicine imaging data, including automated analysis and quantification, within minutes of imaging acquisition. Stroke clinicians now have quick imaging access at their fingertips on their mobile devices with dynamic scrolling, windowing, orientation, and creation of three-dimensional imaging reconstructions. There is no need to log in to PACS to view the images. The team can access images and AI results from anywhere. The AI software serves as a clinical decision support system: It helps clinicians diagnose early ischemic changes, vessel occlusion, and abnormal perfusion. Thus, imaging interpretation need not wait for validation by a radiologist.The AI smartphone applications also make communication easier, with text messaging and calling capabilities to communicate treatment and triage decisions. The on-call stroke team can be notified directly by the AI tool, skipping the radiologist, thus replacing the more traditional workflow where radiologists were the first to interpret and communicate. The coordinated care also can occur between different hospital systems, and multiple team members can participate in real time. The new workflow using AI thus significantly expedites decision-making, patient transfer, and activating the neurointerventional team.Yet, a missing piece in the new AI-powered stroke workflow may be the radiologist. Currently, decisions regarding triage and treatment using imaging data are being made in real time with or without a radiologist's input. More often, the participants in the online communication are the stroke neurologist, emergency physician, neurosurgeon, neurointerventionalist, and stroke nurse coordinator—all discussing the imaging findings and the implications for patient care. Is it an example of the "invisible radiologist," or worse, the "not needed radiologist"? A recent publication from a hub-and-spoke system compared the effect of an AI LVO model on real-world stroke workflow metrics. The time from CT angiography to team notification was 22 minutes shorter, with faster door-to–arterial puncture times for transfer patients (time savings of 23 minutes) with LVO detected by AI versus usual care (27). Another point of interest in this study is that the neuroradiologists documented notification to the stroke team in only 66% of cases, compared with a 100% notification by AI.Radiologists are clearly the imaging experts and well equipped to assist with clinical decision-making, as done routinely in imaging reports, in reading rooms, and during multidisciplinary conferences. Radiologists also have the advantage of viewing the images on high-quality diagnostic monitors, compared with nonradiologists who may use nondiagnostic smartphone screens. AI-enabled decision-making without essential supervision from qualified radiologists can result in medical errors. Automated ASPECTS on noncontrast CT can be inaccurate due to pre-existing white matter disease, prior infarcts, or aneurysm coils (28). Automated LVO tools have challenges in detecting distal occlusions or differentiating between acute versus chronic occlusions and intracranial atherosclerosis (29). Automated perfusion maps can misclassify core ischemic tissue and penumbra due to technical and diagnostic pitfalls. Causes of errors include truncation of time-density curves, partial reperfusion, chronic occlusions, and normal variation in cerebral vasculature (30). Studies have reported a high percentage of misclassification bias in automated CT perfusion analysis (31).Computer-aided diagnosis (CAD) tools such as those for breast cancer detection had disappointing medical outcomes, which pointed out their inherent errors and biases, and made radiologists skeptical. However, those CAD systems were vastly different from the modern AI systems, which utilize newer technology and greater computing power (32,33). The best use case scenario is not autonomous AI, but rather AI partnered with human supervision. It is important to understand the intended clinical use of the technology. The stroke AI algorithms are currently used (and marketed) as an "alert system." Their automated image analysis and volumetrics are not designed to be stand-alone tools. On the other hand, these tools complement the physician to make a timely and accurate diagnosis and treatment decision after integrating the clinical and imaging information. An expert radiologist is a key stakeholder in this process.To change current practices, both academic and community-practice radiologists would need to use mobile devices to partake in conversation with the stroke team to provide the best patient care. Furthermore, interprofessional communication between radiologists and referring clinicians is necessary and appreciated (34), particularly in complex cases or when the AI system has a false-positive or false-negative finding (35). The argument against changing current practice is that radiologists may have to view different algorithms for different disease processes, such as separate applications for code stroke, aortic dissection, pulmonary embolism (36), and fractures (37). Disrupting the current workflow has challenges, especially when high imaging volumes inundate radiology practices. Plus, getting notifications and responding on mobile phones could be considered distractions from being "highly productive." However, it is critical to make contributions in real time at the point of care as an integrated member of the patient care team. Additionally, it is well recognized that active participation in patient care can provide increased satisfaction and break the monotony of interpreting thousands of images in a silo (38).Such wide adoption of AI solutions for code stroke is just the beginning. The evolution is inevitable, particularly in highly time-dependent procedures with many more such algorithms becoming part of clinical practice. AI is a disruptive technology, one with the power to change markets and customers (39). There have been endless debates and discussions on whether AI will replace radiologists. We, as radiologists, assert that AI will not replace us, but rather augment our intelligence and workflow. But when an AI application such as that in code stroke changes clinical practice, are we equipped to adapt? Such changes could mean deviating from the PACS workflow to a mobile application for a particular case and getting notifications directly on phones for time-sensitive critical conditions. Radiologists have an opportunity to situate themselves at the center of patient care but will require transforming the operational framework and rebranding ourselves. Otherwise, we may lose the potential to add value where it matters.It is an exciting time to be a radiologist, a time for a culture change. The most significant skill needed is adaptability. Are we as radiologists ready to embrace these new AI models of care and redefine our role? Are we willing to shift from the "invisible radiologist" to being the "embedded radiologist"? (40).Disclosures of Conflicts of Interest: A.V. Payment for all research grants made to the institution: NIH/NINDS NS103824. NINDS/NIA NS117643, NIH/NINDS NS100417, NIH/NINDS 1U01NS100699, NIH/NINDS, U01NS110772 PI, Imaging Core Lab, ENDOLOW Trial, Cerenovus; guest editor for March 2022 issue of Radiology: Artificial Intelligence issue. L.S. No relevant relationships.Authors reported no funding for this work.References1. GBD 2016 Neurology Collaborators.Global, regional, and national burden of neurologic disorders, 1990–2016: a systematic analysis for the Global Burden of Disease study 2016. Lancet Neurol 2019;18(5):459–480. Crossref, Medline, Google Scholar2. Berkhemer OA, Fransen PSS, Beumer D, et al. A randomized trial of intraarterial treatment for acute ischemic stroke. N Engl J Med 2015;372(1): 11–20. Crossref, Medline, Google Scholar3. Campbell BCV, Mitchell PJ, Kleinig TJ, et al. Endovascular therapy for ischemic stroke with perfusion-imaging selection. N Engl J Med 2015; 372(11):1009–1018. Crossref, Medline, Google Scholar4. Albers GW, Marks MP, Kemp S, et al. Thrombectomy for Stroke at 6 to 16 Hours with Selection by Perfusion Imaging. N Engl J Med 2018; 378(8):708– 718. Crossref, Medline, Google Scholar5. Saver JL, Goyal M, Bonafe A, et al. Stent-retriever thrombectomy after intravenous t-PA vs. t-PA alone in stroke. N Engl J Med 2015; 372(24):2285– 2295. Crossref, Medline, Google Scholar6. Goyal M, Demchuk AM, Menon BK, et al. Randomized assessment of rapid endovascular treatment of ischemic stroke. N Engl J Med 2015; 372(11):1019– 1030. Crossref, Medline, Google Scholar7. Jovin TG, Chamorro A, Cobo E, et al. Thrombectomy within 8 hours after symptom onset in ischemic stroke. N Engl J Med 2015; 372(24):2296– 2306. Crossref, Medline, Google Scholar8. Nogueira RG, Jadhav AP, Haussen DC, et al. Thrombectomy 6 to 24 Hours after Stroke with a Mismatch between Deficit and Infarct. N Engl J Med 2018;378(1): 11–21. Crossref, Medline, Google Scholar9. Hill MD, Hachinski V. Stroke treatment: time is brain. Lancet 1998; 352(Suppl 3):SIII10– SIII14. Crossref, Medline, Google Scholar10. Kunz WG, Hunink MG, Almekhlafi MA, et al. Public health and cost consequences of time delays to thrombectomy for acute ischemic stroke. Neurology 2020;95(18): e2465–e2475. Crossref, Medline, Google Scholar11. Meretoja A, Keshtkaran M, Tatlisumak T, Donnan GA, Churilov L. Endovascular therapy for ischemic stroke: Save a minute-save a week. Neurology 2017; 88(22):2123– 2127. Crossref, Medline, Google Scholar12. Soun JE, Chow DS, Nagamine M, et al. Artificial Intelligence and Acute Stroke Imaging. AJNR Am J Neuroradiol 2021; 42(1):2– 11. Crossref, Medline, Google Scholar13. Barber PA, Demchuk AM, Zhang J, Buchan AM. 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 2000;355( 9216):1670–1674. Google Scholar14. Nagel S, Sinha D, 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(6):615–622. Crossref, Medline, Google Scholar15. Maegerlein C, Fischer J, Mönch S, et al. Automated calculation of the Alberta stroke program early CT score: Feasibility and reliability. Radiology 2019;291(1):141– 148. Link, Google Scholar16. Bouslama M, Ravindran K, 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( 2):634–641. Crossref, Medline, Google Scholar17. Morey JR, Zhang X, Yaeger KA, et al. Real-World Experience with Artificial Intelligence-Based Triage in Transferred Large Vessel Occlusion Stroke Patients. Cerebrovasc Dis 2021;50(4):450–455. Crossref, Medline, Google Scholar18. Powers WJ, Rabinstein AA, 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( 12):e344–e418. Crossref, Medline, Google Scholar19. Hassan AE. New Technology Add-On Payment (NTAP) for Viz LVO: a win for stroke care. J Neurointerv Surg 2021;13(5):406–408. Crossref, Medline, Google Scholar20. Al-Kawaz M, Primiani C, Urrutia V, Hui F. Impact of RapidAI mobile application on treatment times in patients with large vessel occlusion. J Neurointerv Surg 2021. https://doi.org/10.1136/neurintsurg-2021-017365. Published online April 1, 2021. Crossref, Medline, Google Scholar21. Adhya J, Li C, Eisenmenger L, et al. Positive predictive value and stroke workflow outcomes using automated vessel density (RAPID-CTA) in stroke patients: One year experience. Neuroradiol J 2021; 34( 5): 476– 481. Crossref, Medline, Google Scholar22. Hassan AE, Ringheanu VM, Rabah RR, Preston L, Tekle WG, Qureshi AI. Early experience utilizing artificial intelligence shows significant reduction in transfer times and length of stay in a hub and spoke model. Interv Neuroradiol 2020; 26( 5): 615– 622. Crossref, Medline, Google Scholar23. van Leeuwen KG, Meijer FJA, Schalekamp S, et al. Cost-effectiveness of artificial intelligence aided vessel occlusion detection in acute stroke: an early health technology assessment. Insights Imaging 2021; 12( 1): 133. Crossref, Medline, Google Scholar24. Brainomix. https://www.brainomix.com/. Accessed November 28, 2021. Google Scholar25. RAPIDAI. Pushing boundaries in stroke patient care. https://www.rapidai.com/stroke. Accessed November 28, 2021. Google Scholar26. Viz.ai. https://www.viz.ai/. Accessed November 28, 2021. Google Scholar27. Elijovich L, Dornbos Iii D, Nickele C, et al. Automated emergent large vessel occlusion detection by artificial intelligence improves stroke workflow in a hub and spoke stroke system of care. J Neurointerv Surg 2021. https://doi.org/10.1136/neurintsurg-2021-017714. Published online August 20, 2021. Crossref, Google Scholar28. Guberina N, Dietrich U, 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( 9): 889– 901. Crossref, Medline, Google Scholar29. Chatterjee A, Somayaji NR, Kabakis IM. Abstract WMP16: Artificial Intelligence Detection of Cerebrovascular Large Vessel Occlusion-Nine Month, 650 Patient Evaluation of the Diagnostic Accuracy and Performance of the Viz.ai LVO Algorithm.Stroke 2019;50(Suppl_1). Google Scholar30. Vagal A, Wintermark M, Nael K, et al. Automated CT perfusion imaging for acute ischemic stroke: Pearls and pitfalls for real-world use. Neurology 2019; 93( 20): 888– 898. Crossref, Medline, Google Scholar31. Geuskens RREG, Borst J, Lucas M, et al. Characteristics of Misclassified CT Perfusion Ischemic Core in Patients with Acute Ischemic Stroke. PLoS One 2015; 10( 11): e0141571. Crossref, Medline, Google Scholar32. Oakden-Rayner L. The Rebirth of CAD: How Is Modern AI Different from the CAD We Know? Radiol Artif Intell 2019; 1( 3): e180089. Link, Google Scholar33. Kohli A, Jha S. Why CAD Failed in Mammography. J Am Coll Radiol 2018; 15( 3 Pt B): 535– 537. Crossref, Medline, Google Scholar34. Fatahi N, Krupic F, Hellström M. Difficulties and possibilities in communication between referring clinicians and radiologists: perspective of clinicians. J Multidiscip Healthc 2019; 12( 555): 564. Google Scholar35. Liebeskind DS. Artificial intelligence in stroke care: Deep learning or superficial insight? EBioMedicine 2018; 35( 14): 15. Google Scholar36. Weikert T, Winkel DJ, Bremerich J, et al. Automated detection of pulmonary embolism in CT pulmonary angiograms using an AI-powered algorithm. Eur Radiol 2020; 30( 12): 6545– 6553. Crossref, Medline, Google Scholar37. Voter AF, Larson ME, Garrett JW, Yu JJ. Diagnostic Accuracy and Failure Mode Analysis of a Deep Learning Algorithm for the Detection of Cervical Spine Fractures. AJNR Am J Neuroradiol 2021; 42( 8): 1550– 1556. Crossref, Medline, Google Scholar38. Friedberg E, Chong ST, Pyatt RS Jr, et al. Unifying the Silos of Subspecialized Radiology: The Essential Role of the General Radiologist. J Am Coll Radiol 2018; 15( 8): 1158– 1163. Crossref, Medline, Google Scholar39. Ratner M. FDA backs clinician-free AI imaging diagnostic tools. Nat Biotechnol 2018; 36( 8): 673– 674. Crossref, Medline, Google Scholar40. Gunderman RB, Chou HY. The Future of Radiology Consultation. Radiology 2016; 281( 1): 6– 9. Link, Google ScholarArticle HistoryReceived: July 19 2021Revision requested: Aug 6 2021Revision received: Dec 27 2021Accepted: Dec 29 2021Published online: Jan 19 2022 FiguresReferencesRelatedDetailsCited ByRecruitment in Acute Stroke Trials: Challenges and Potential SolutionsJoseph P.Broderick, Yasmin N.Aziz, Opeolu M.Adeoye, James C.Grotta, Andrew M.Naidech, Andrew D.Barreto, Colin P.Derdeyn, Heidi J.Sucharew, Jordan J.Elm, PoojaKhatri2023 | Stroke, Vol. 54, No. 2Randomized Clinical Trials of Artificial Intelligence in Medicine: Why, When, and How?Seong HoPark, Joon-IlChoi, LaureFournier, BaptisteVasey2022 | Korean Journal of Radiology, Vol. 23, No. 12Recommended Articles Detection of Acute Infarction on Non–Contrast-enhanced CT: Closing the Gap with MRI via Machine LearningRadiology2020Volume: 294Issue: 3pp. 645-646Imaging-based Selection for Endovascular Treatment in StrokeRadioGraphics2019Volume: 39Issue: 6pp. 1696-1713Recent Endovascular Trials: Implications for Radiology Departments, Radiology Residency, and Neuroradiology Fellowship Training at Comprehensive Stroke CentersRadiology2016Volume: 278Issue: 3pp. 642-645Symmetric CTA Collaterals Identify Patients with Slow-progressing Stroke Likely to Benefit from Late ThrombectomyRadiology2021Volume: 302Issue: 2pp. 400-407Attenuation Changes in ASPECTS Regions: A Surrogate for CT Perfusion–based Ischemic Core in Acute Ischemic StrokeRadiology2019Volume: 291Issue: 2pp. 451-458See More RSNA Education Exhibits Acute Ischemic Stroke: Time is Life — An Approach for No MistakesDigital Posters2019Detection of Ischemic Core: Technological of Noncontrast CT and Image Analysis of Contrast CTDigital Posters2018Interpretation of CT Perfusion in Acute Ischemic Stroke: A Practical GuideDigital Posters2019 RSNA Case Collection Acute ischemic strokeRSNA Case Collection2020Ischemic closed loop small bowel obstruction RSNA Case Collection2020Acute bilateral medial medullary infarctRSNA Case Collection2021 Vol. 4, No. 2 Metrics Downloaded 1,698 times Altmetric Score PDF download
Referência(s)