The Growing Role of AI in Healthcare
Consider how often most people think about their health – and, by extension, about the care available to them during a healthcare event that requires professional treatment. Compounded by healthcare’s position as a hot-button issue across the entire sociopolitical spectrum, notably (but far from exclusively) in the U.S. over the course of the past decade, it can seem as if the topic hardly ever leaves people’s minds.
“AI is poised to play a major role in the global healthcare industry.”
The degree to which healthcare matters to many segments of the business world – in some cases as a market for their products or services, in others as a matter to be managed in the form of employee benefits – is similarly and undeniably significant. It’s also become increasingly clear that the proving grounds for the latest and greatest tech advances, including artificial intelligence, will be within the greater context of the health sector, among others. Artificial Intelligence (AI) in particular may hold the keys to numerous breakthroughs in the field. As such, everyone from players directly involved in the industry to companies overhauling their process management software should pay close attention to all of the most significant developments in this high-priority tech segment.
AI suitable to the market’s size
According to Forbes, the total cost of healthcare around the world will, in the not-too-distant future, account for at least 10 percent of most nations’ entire gross domestic product, reaching a sum of more than $9 trillion. It may not be all that long before that figure reaches $9.5 or $10 trillion, given the direction in which costs of living are trending throughout many of the world’s biggest, most economically sound countries.
Adding further difficulties to the matter is the expected deficit of medical professionals in numerous subcategories, such as the long-projected nursing shortage expected by the American Association of Colleges of Nursing, that will begin affecting the U.S. as senior citizens start eclipsing all other demographics of the American population. Less developed countries, meanwhile, could experience shortfalls of many major types of healthcare workers – doctors, nurses, administrators and support staff. When considering the broad complexity of these potential – and, in some cases, all but inevitable – healthcare issues, it becomes clear that AI’s vast capabilities can aid medicine immensely: for risk identification, improved self-management in telehealth settings and potential bottom-line cost reductions over time. AI may perhaps be more broadly applicable across healthcare than any other similarly cutting-edge technology.
Major diagnostic value
Diagnostics will likely be the most impactful application of AI in healthcare. At the behest of the U.S. Department of Health & Human Services, the elite JASON scientific advisory group prepared a comprehensive December 2017 report outlining major examples where high-level automation could improve medical diagnoses through earlier detection of certain diseases. Skin cancer stood out notably among these because the rarity of melanomas (responsible for about 75 percent of skin-cancer fatalities) complicates their early identification even by talented dermatologists. Convolutional neural network algorithms detected malignant melanomas at rates equal to or slightly greater than dermatologists not using AI.
JASON’s report also highlighted AI’s success in more effective identification of coronary artery disease by using non-invasive fluid flow reserve analysis. Eliminating invasive coronary angiography could save time for both patients and medical staff while also ensuring those suffering from CAD begin receiving appropriate treatment more quickly.
Similarly, a study in the British Medical Journals’ Stroke and Vascular Neurology publication found immense potential in AI for predicting the likelihood of stroke, in addition to properly diagnosing the condition when it manifests and assessing the efficacy of relevant treatments. Machine-learning algorithms applied in conjunction with support vector machines helped lead to this discovery: In one experiment, SVMs showed an accuracy factor of nearly 88 percent in identifying patients suffering from strokes.
Assistance with multiple tasks to increase efficiency
It will be some time before AI-dominated diagnostics become the norm across the board. However, the technology’s current iterations are eminently capable of assisting medical professionals in a number of important contexts. According to a separate Forbes report, workflow and administrative support will perhaps be the easiest form of AI usage for medical facilities to implement and successfully leverage, with the potential to facilitate $18 billion in sector-wide savings. Notable AI-based tasks include data mining to develop personalized treatment plans based on extant information from past and present medical research documents, as well as high-end voice-to-text transcription, support for better robotic surgery and virtual nursing assistants, to name just a few.
FG-AI4H aims to address a notable AI limitation
May 15, 2018, marked the start of the second annual AI Good for Global Summit, overseen by the International Telecommunications Union, a subsidiary of the UN focused on improving the global greater good by using high-end technological developments to bolster worldwide connectivity. This led to the creation of the Focus Group on AI for Health two months later as a collaboration with the World Health Organization, and the AI-dedicated task force had its first meeting Sept. 25-27 in Geneva. A worldwide “Call for Proposals” came about a week after the FG-AI4H initially convened, soliciting accounts of use cases in which AI supports clinical or public health operations.
Standardized practices applied worldwide could greatly bolster AI’s efficacy in healthcare.
The ITU-WHO joint task force is well aware that AI, in its current iterations, isn’t a flawless tool for any application, medical or otherwise. As such, ensuring that certain industry-wide standards are put in place for those looking to create AI solutions for healthcare applications, to help overcome such limitations, is among FG-AI4H’s most notable goals for the near future. Thomas Wiegand, a professor at the Technical University of Berlin, the executive director of the Fraunhofer Heinrich Hertz Institute and chairperson of FG-AI4H, addressed this in a video interview when asked about challenges to worldwide AI healthcare implementations:
“We are seeing mostly limited projects where you have a given dataset and you develop your algorithm for it and then you report on what comes out,” Wiegand said. “The problem with that is, how does this generalize to worldwide data that are coming in for this algorithm? We don’t know. And how does a regulator … make a recommendation on a particular algorithm when they don’t know how well it performs in a general sense?”
Wiegand went on to explain FG-AI4H’s intent to create global guidelines for approaching health problems using AI – similar to those developed by ISO and other standardization groups – as well as a broadly applicable rubric for assessing AI projects’ performances.
The future is unwritten – and rife with possibility
It’s important to note that the day when AI is a dominant feature of the healthcare industry – or any other sector, for that matter – is still a ways off in the future. It’ll be vital for governments to regulate medical AI and prevent its misuse. That being said, its current uses are nonetheless exciting and valuable. Hospital networks can bolster the effectiveness of their existing automation by using Appian’s Application Platform to design AI oversight apps for staff, as well as patient portals to improve their care experience.