The role of artificial intelligence (AI) is changing in many sectors of society, including medicine and health care. It has the potential to change the way physical therapy is practiced. This article is based on a 2020 Annual Meeting Webinar presented by Anthony Chang.
A machine can now read human tissue, process it, and predict the likelihood of it being a tumor in 100 seconds. In medical centers where available, this new technology has completely changed the landscape for surgeons who need tissue biopsies.
We have had evidence-based medicine for the last few decades, along with a move toward precision medicine and population health. Now, we are moving toward intelligence-based medicine: a combined effort between artificial intelligence and human cognition.
In the 1940s, Alan Turing was the first to come up with the concept of using a machine to deal with massive amounts of data that humans simply cannot handle. In 1956, John McCarthy coined the term "Artificial Intelligence." In the 1990s, a chess champion lost to a computer. More recently, in 2011, the IBM supercomputer Watson beat the human champions in Jeopardy! Just a few years ago, Google's DeepMind was able to beat the human champion in the game AlphaGo, which is more complicated than chess.
We are entering a new era with AI, and we are lucky that this enormous resource is coming ahead of schedule. There have been three areas feeding into this growth:
Previously, humans enjoyed an edge over AI thinking and performance. However, we are entering an era where AI has started to match and exceed human performance in some particular tasks, such as conversational speech recognition, driving vehicles, playing certain games, and classifying skin cancer. Looking into the future, we will move into an era where general AI exceeds human performance and reasoning in complex tasks, including writing best-selling novels and performing surgery. This is a partnership. AI can help us, as humans, improve our intelligence and learn. AI functions at three levels: assisted, augmented, and autonomous.
Level of Human Involvement
Example in Healthcare
System providing and automating repetitive tasks
Little or none
UR robots for bloodwork (Copenhagen Hospital)
Humans and machines collaboratively make decisions
Some or high
Watson for Oncology (Memorial Sloan Kettering)
Decisions made by adaptive intelligent systems autonomously
IDxDR for retinal images (University of Iowa)
AI has several components that are important to understand. Within AI, there is both machine learning and, within that, deep learning. With deep learning, computers can take anything we give them to adjust their performance and behavior accordingly. Data science, and within this data analytics and data mining, feeds into that learning. And, finally, mathematics, which includes statistics, also plays an important role.
Convolutional Neural Network (CNN) is a form of deep learning to help with image interpretation. This involves hundreds of layers that essentially mimic how the cortex interprets light and images. Usually, AI can interpret images far faster and far more accurately. The error rate dropped significantly since 2010, which is why most AI technologies surpass humans at this point. However, there are some issues and weaknesses. For example, computers have a hard time differentiating blueberry muffins and Chihuahuas. Even small children can differentiate these two items easily, so, obviously, we are missing something between how small children learn with a small amount of data versus how computers learn with massive amounts of data.
We used to pit AI against humans. However, with advances in AI, this comparison is no longer needed. We have moved past that and are instead looking for synergy between AI and humans. For physical therapy, digital biofeedback tools could help feed diagnostics and therapy, especially in telehealth settings. Instead of choosing between only in-person and only biofeedback, combining the two can help us achieve the best results.
Currently, the hype around AI is still exceeding capability. However, ten years from now, we may look back and say we did not expect enough from AI—we will see how much of a game-changer it is across health care. One analogy to consider is the biomimicry of flight. It took the Wright brothers many years to build a working plane, but once they achieved flight, aeronautical technology quickly advanced. As Bill Gates said, "We always overestimate the change that will occur in the next two years and underestimate the change that will occur in the next ten. Don't let yourself be lulled into inaction." We cannot wait for the advances to come; we need to work toward them.
These advances may make it additionally frustrating as we live in a pandemic that is such a profound issue. How can we launch something into space and have it come back to an exact spot on Earth but not figure out a pandemic? Well, the answer is simple. Engineering feats involve a series of very complicated facts that, once deconstructed, can be solved with a relatively high level of precision. However, a pandemic involves human behavior, which has a very high level of unpredictability. Any human behavior or biological phenomenon is complex.
Still, technology and AI can help us address complex issues. In the medical field, let us not place human intelligence and machine intelligence in separate realms. Instead, we should combine them to make "Medical Intelligence." By thinking of it that way, AI can be a paradigm shift in health care and medicine.
Anthony Chang, MD, MBA, MPH MS, is a pediatric cardiologist and Medical Director of the Heart Failure Program at CHOC Children's. He is the Chief Intelligence and Innovation Officer and Founder and Medical Director of The Sharon Disney Lund Medical Intelligence and Innovation Institute (MI3) at CHOC and serves as co-director of the MI3 Summer Internship Program. He is the founder of Pediatrics 2040 and the founder and chair of the International Society of Pediatric Innovation (iSPI), as well as chairman and founder of Artificial Intelligence in Medicine (AIMed). Dr. Chang also launched the Medical Intelligence Society (MIS) and is the founder of Medical Intelligence 10 (MI10). He is the author of Intelligence-Based Medicine: Data Science, Artificial Intelligence, and Human Cognition in Clinical Medicine and Healthcare, which was published in July 2020.
Copyright ©2021 All Rights Reserved.