By: Seethal Sara Thomas, FNP-BC
Tools are a driving force in the evolution and efficiency of humankind. Tools helped humans access new food sources, defend themselves, and adapt to changing environments. One tool at the forefront of innovation is Artificial Intelligence (AI). Accessing AI features is as easy as typing words in an address and search bar. Integrating AI into workflows and using it for specialized purposes requires astute implementation. This is crucial particularly when using tools in decision making within healthcare. A subset of AI called machine learning (ML) is used to teach machines how to extract relevant data with more efficiency. The use of ML in neurology, specifically in stroke detection and treatment has been studied over the past few years. Pattern recognition with high contrast colors in brain imaging is at the core of detecting stroke pathology on computed tomography (CT). A rapid, accurate, and continuous scanner able to communicate with all members of the stroke team can assist in timely diagnosis and treatment (Shlobin et al., 2022). Large vessel occlusions (LVOs) have a better outcome the faster they are detected and treated.
ML detection in the acute stroke setting is useful for bringing concerning imaging to clinician attention and communicating within the stroke team and having a centralized place for clinicians on different systems to access the same imaging. This online data repository of pathologic cases can assist with clinician research and student teaching. A systemic review of studies including ML algorithms with existing AI applications for LVO strokes showed to be accurate in identifying LVO on computed tomography (Shlobin et al., 2022). When properly designed, the ML software has the ability to improve triage of candidates, especially candidates for endovascular mechanical thrombectomy (Olive-Gadea et al., 2020). Having this capability to distinguish LVOs reliably and rapidly can impact the systems and efficiency with which patients with strokes are transferred to different hospitals. Not all hospitals are equipped with computed tomographic angiogram (CTA) access 24/7. If the ML can detect LVO candidates from just a non-contrast computerized tomography (NCCT), this can help navigate the patient to a thrombectomy capable center even if the ability to obtain CTA is not available. Decisions about transfer should consider patient history, functional status, and clinical exam. Clinician insight remains important as false positives are possible (Olive-Gadea et al., 2020). Despite the potential for false positives, transferring the patient to a more capable stroke site generally allows for better care if the patient’s clinical condition worsens.
The true benefit of outcomes and cost savings in integrating state of the art technology is seen when workflows and proper systems are in place. Acute ischemic stroke treatment has international guidelines in place for acute treatment. There are research efforts to define and fine tune the workflow of acute ischemic stroke treatment to improve patient outcomes (Goyal et al., 2021). Cost drivers of stroke treatment include the diagnostic portion which is the imaging and the treatment portion which are endovascular thrombectomy and intravenous thrombolysis if the patient is an appropriate candidate. Endovascular thrombectomy is a main hospital cost driver but has shown to be partially offset by decreased in median length of stay and in-hospital mortality (Christensen et al., 2022). The improved outcomes of treatment of acute ischemic stroke of appropriate candidates are a feat of modern medicine that should be celebrated. There are other areas of medicine in which AI digital scanning are being investigated, but in acute stroke triage, the focus on speed and accuracy makes this an area of particular interest. Improved access to acute ischemic stroke treatment and using AI as a tool used within a reliable workflow may be the next step in improving and increasing access to acute ischemic stroke care.
References
Christensen, E. W., Pelzl, C. E., Hemingway, J., Wang, J. J., Sanmartin, M. X., Naidich, J. J., Rula, E. Y., & Sanelli, P. C. (2022). Drivers of Ischemic Stroke Hospital Cost Trends Among Older Adults in the United States. Journal of the American College of Radiology. https://doi.org/10.1016/j.jacr.2022.09.026
Goyal, M., Saver, J. L., Ganesh, A., McDonough, R. V., Roos, Y. B. W. E. M., Boulouis, G., Kurz, M., Psychogios, M., Holmin, S., Majoie, C. B. L. M., Bourcier, R., Chandra, R., Yoshimura, S., Yavagal, D., Gory, B., Taschner, C., Buck, B., Jadhav, A., Hill, M. D., & Ospel, J. M. (2021). Standardized Reporting of Workflow Metrics in Acute Ischemic Stroke Treatment: Why and How? Stroke: Vascular and Interventional Neurology, 1(1). https://doi.org/10.1161/svin.121.000177
Olive-Gadea, M., Crespo, C., Granes, C., Hernandez-Perez, M., Pérez de la Ossa, N., Laredo, C., Urra, X., Carlos Soler, J., Soler, A., Puyalto, P., Cuadras, P., Marti, C., & Ribo, M. (2020). Deep Learning Based Software to Identify Large Vessel Occlusion on Noncontrast Computed Tomography. Stroke, 51(10), 3133–3137. https://doi.org/10.1161/strokeaha.120.030326
Shlobin, N. A., Baig, A. A., Waqas, M., Patel, T. R., Dossani, R. H., Wilson, M., Cappuzzo, J. M., Siddiqui, A. H., Tutino, V. M., & Levy, E. I. (2022). Artificial Intelligence for Large- Vessel Occlusion Stroke: A Systematic Review. World neurosurgery, 159, 207–220.e1. https://doi.org/10.1016/j.wneu.2021.12.004
Assessed and Endorsed by the MedReport Medical Review Board