In the last several weeks, FDA Commissioner Scott Gottlieb has commented on artificial intelligence and the future of medicine. For nearly a decade, many visionaries and innovators in medicine have believed this to be the case. Finally, the FDA is getting on board by creating an internal data science incubator, called Information Exchange and Data Transformation, or INFORMED. Already this year, the FDA has approved a clinical support software that uses AI algorithms to detect changes in brain function that can alert neurologists much faster than traditional technologies. In addition, the agency also approved an AI-based tool that uses a camera to detect diabetic retinopathy — a condition that can lead to blindness if not detected and treated early. Gottlieb commented at the Health Datapalooza conference in D.C. last month that the FDA was working on new regulatory guidelines that will allow the agency to keep up with new AI-based technologies and “promote innovation in this space.”
What is artificial intelligence, and why does it matter?
Artificial intelligence, machine learning, and deep learning are the buzzwords that we are beginning to hear in all types of industry — including medicine. While all of these terms refer to AI, they are not all the same thing.
The easiest way to think about the relationship of AI to machine learning and deep learning is to think of them as a group of Russian nesting dolls or Matryoshka Dolls. At the center of the stack of dolls is deep learning, followed by machine learning, and the final and largest outer doll is AI. In the simplest of terms, AI is the simulation of human intelligence by computer systems. The concept of AI has been around since the early 1950s but has seen rapid development in the last decade. AI allows machines to perform tasks that normally require human intelligence such as visual perception, speech recognition, and decision-making.
Machine learning uses algorithms to analyze data, learn from it, and ultimately make decisions, determinations, and predictions. In short, machine learning occurs when a computer actually evolves and “learns” without specific programming via complex pattern recognition. With machine learning, a computer is able to actually modify “itself” when exposed to new data. Machine learning is a powerful tool that can be applied in many medical settings to minimize errors and maximize the accuracy of a prediction (or diagnosis).
Deep learning is a subset of machine learning (and thus the smallest doll in the stack). In deep learning algorithms, artificial neural networks are used to recognize patterns and cluster and classify data. Deep learning maps inputs to outputs and finds correlations. This is exactly what physicians do when making treatment decisions for a particular disease. In addition, deep learning relies on “reinforcement centers” where a machine is able to actually obtain a complex objective or goal (such as making a correct diagnosis or choosing a correct treatment). Ultimately, in deep learning, a machine learns the concept of “delayed gratification” and is able to correlate immediate actions with results that may be realized weeks to months later.
How can AI impact medicine?
As you might imagine, the applications in medicine for these tools are limitless and may ultimately result in better treatments and cures for chronic disease through pattern recognition and early diagnosis and intervention. AI and similar technologies can speed diagnosis and early detection of disease. The ability to make real-time decisions and engage patients via mobile devices will likely result in improved outcomes. Ultimately, the use of AI-equipped medical applications and tools on smartphones and other mobile devices could result in a substantial cost savings for the healthcare system by obviating the need for expensive specialists and treatments (thru early detection and intervention). AI will allow us to catch very serious diseases much earlier in their course and will prevent many common complications that contribute to rising healthcare costs.
Ultimately, AI will make personalized medicine — the use of individual health data paired with predictive analytics — more accessible for all patients. Personalized medicine, as described by Dr. Eric Topol, will allow for the prescription of better therapies that are catered to an individual’s own biology. Rather than basing treatments for one person on an entire population, personalized medicine (which can be facilitated by AI), will base your treatment on your very own genetics.
AI is already being used in medicine: From cameras that can diagnose eye problems related to diabetes to the Alive Cor device that can record and interpret ECGs from your phone to automated remote follow-ups of implanted pacemakers and implantable defibrillators and other cardiac devices via a software offered by PaceMate, AI is already making impacts in the lives of everyday patients.
What do we need from the FDA in order to move forward with AI in medicine?
The FDA approval process can be arduous — it is often expensive, tedious, and time-consuming. Treatments are often delayed due to drawn-out approval processes.
For example, in the past, whenever a medical device is approved and a software update or change is made, another FDA approval process must be undertaken. These antiquated approval processes can result in months of delays, and many patients may be inconvenienced by FDA regulations. While the FDA is making strides towards streamlining approvals through the passage of the Obama-era 21st Century Cures Act, little has been done to address newer technologies such as artificial intelligence.
Gottlieb’s endorsement of AI in his recent speech gives me hope that the agency will continue to embrace new innovations that will potentially change the way medical care is delivered. The FDA has already created an “incubator” to work on the development of new digital health technologies, and Gottlieb has promised more actions to promote the development of AI applications in medicine. We must continue to demand that our FDA streamline approvals for AI-related devices and work with entrepreneurs, researchers, and computer scientists to bring these much-needed technologies to patients today.
Kevin Campbell (@DrKevinCampbell) is a contributor to the Washington Examiner’s Beltway Confidential blog. He is an internationally-recognized cardiologist and medical, health, and wellness expert. He has authored two books and appears regularly on Fox News, Fox Business, CBS and other media outlets. Dr. Campbell is the CEO of PaceMate, a healthcare data solutions company.
