While healthcare artificial intelligence (AI) has experienced triumphs—ranging from applications in stroke care and oncology, to hospital management and day-to-day tools for clinicians—it has not always lived up to its initial hype and promises.
Case in point: The January 2022 announcement that a private equity group would acquire much of IBM Watson Health marked a drastic shift for IBM, which had previously predicted that it would transform the healthcare analytics industry. IBM said that it “remains committed to Watson, our broader AI business, and to the clients and partners we support in healthcare IT,” but the sale came amid reports of IBM Watson Health profitability concerns. Some also say a Watson oncology application has been slow to learn; it has had tepid adaptation by hospitals. For many, this raises questions about the broader implications of AI in healthcare.
AI is hungry for quality data
In a separate AI scenario, a report by the Alan Turing Institute in the UK showed that medical and other experts were constrained in their ability to access a wide range of different types of data during the coronavirus pandemic, with the report complaining that “the single most consistent message across the workshops was the importance—and at times lack—of robust and timely data.”
Scientists and others experienced problems with locating available data, as well as accessing and standardizing it for use, which are all requirements for accurate machine learning. The data woes were not limited to those associated with COVID-19 hospitalizations early in the pandemic, but extended to areas such as the effectiveness of mask-wearing and lockdowns, as well as localized employment and productivity information.
Applications for COVID-19
A notable success story: Researchers at the Regenstrief Institute and Indiana University were able to demonstrate that it is feasible that machine learning models could leverage public health data to determine the likelihood that a COVID-19 patient would require healthcare resources.
Overall, perhaps the question of whether healthcare AI is living up to its initial promises should be more nuanced: A 2017 review in Stroke and Vascular Neurology forecast that AI is not set to replace human physicians anytime soon, but that AI can assist them in making decisions and that it may “replace human judgment in certain functional area of healthcare (eg, radiology).”
These predictions seem to be largely holding true, here in 2022: Instead of rendering clinicians obsolete, AI often functions alongside them by analyzing enormous amounts of data in seconds, enabling healthcare professionals to focus on treatment decisions.
Smaller healthcare AI successes
Many healthcare AI success stories appear not to have had much “hype” to begin with: Along with an array of other startups, AI has been leveraged by companies such as Viz.ai, whose Intelligent Care Coordination Platform is being used by emergency room physicians to rapidly upload stroke patients’ bloodwork and CT scans to smartphones, enabling neurosurgeons to immediately treat patients.
In 2021, Viz.ai announced new data supporting the use of its technology during emergencies, noting the “data demonstrated reduced door-in door-out times at primary stroke centers (PSCs), improved door-in-to-puncture times at comprehensive stroke centers (CSCs), and improved reperfusion rates.”
In another smaller-scale example, AI has been used by researchers at the UK’s University of Leeds to predict heart attacks by, in effect, comparing any damage in the minuscule blood vessels found in patients’ retinas against a large retinal database. Resulting vascular-disease predictions were successful.
With an outcome that arguably showcases AI’s wide-ranging potential, the University of Leeds authors wrote in Nature: “Our results indicate that one could identify patients at high risk of future myocardial infarction from retinal imaging available in every optician and eye clinic.”
The Alan Turing Institute. Data Science and AI in the Age of COVID-19Data
Amisha, Malik P, Pathania M, Rathaur VK. Overview of artificial intelligence in medicine. J Family Med Prim Care. 2019;8(7):2328-2331.
Diaz-Pinto A, Ravikumar N, Attar R, et al. Predicting myocardial infarction through retinal scans and minimal personal information. Nat Mach Intell. 2022;4(1):55-61.
Francisco partners to acquire ibm’s healthcare data and analytics assets. IBM Newsroom.
Jiang F, Jiang Y, Zhi H, et al. Artificial intelligence in healthcare: past, present and future. Stroke Vasc Neurol. 2017;2(4):230-243.