Whenever we hear the term “artificial intelligence” (AI), we imagine robots taking our jobs, making us obsolete. As AI-powered computers are programmed to make decisions with minimal human input, some wonder if they will soon replace our doctors in making difficult decisions.
It’s important to separate fact from science fiction, says David B. Agus, MD, a professor of medicine and engineering at the University of Southern California Keck School of Medicine and Viterbi School of Engineering.
In the healthcare field, Artificial Intelligence refers more to hospitals and doctors accessing vast data sets of potentially life-saving information rather than robotics. Patients, geographical locations, and innumerable health conditions are all included in this data, which includes the treatment methods used and their outcomes, survival rates, and speed of care. Through the use of new computing power, we are able to detect and analyze large and small trends from the data, as well as make predictions through machine learning that identify potential health outcomes.
In machine learning, statistical techniques are used to provide computer systems with the ability to learn from incoming data, identify patterns, and make decisions without direct human intervention.
By using targeted analytics, doctors may be better able to assess risk, make the correct diagnosis, and offer patients more effective treatments, says Agus, author of The Lucky Years: How to Thrive in the Brave New World of Health and The End of Illness. AI’s potential to improve health care is “staggering” to him.
“We have a lot of data that we have been collecting over decades,” he says. “Computer power enables us to use data in a way that benefits patients for the first time.”
It is challenging to glean machine learning insights when millions of patients share millions of healthcare data points, he says, because of how disparate these data are. Furthermore, for machines to be effective, these data have to be collected appropriately.
He gives an example. “A study came out recently that showed that if you have ovarian cancer, and you happen to also be on a beta-blocker — a drug that [can be] used for blood pressure — you lived four-and-a-half years longer,” he says. This is an observation we wouldn’t have made through biology. Big data shows us this. Now [this finding] needs to go to a big trial to see if it’s true.”
From the patient’s perspective, “the most exciting thing is that [AI] allows doctors to personalize care, something we’ve dreamed of for decades,” he claims.
In Agus, a patient’s symptoms are instantly matched to patients who have similar symptoms. “I pull them out of a database,” he says, “and I can say, ‘Here are their reactions.’ Machine learning and AI allow me to [access] all of the information and have a very educated discussion with the patient” sitting in the exam room, “unlocking data [on health conditions] that historically we’ve made simple decisions about. AI allows us to get much deeper and look for associations the human brain cannot do – but a computer can.”
The use of analytics in healthcare is not without critics, but most of the criticism is focused on how big data can be used to evaluate, reward, and penalize a hospital — or even an individual surgeon.
According to Jerry Muller, author of 2018’s The Tyranny of Metrics, such measurements can impact how, when, and even whether a patient is treated. “Nowhere are metrics more popular than in medicine,” he says. The stakes are high, he concludes, with lives on the line.
The problem, Muller says, is human nature, which is known for “gaming” the numbers for survival.
The author points out that in-demand surgeons who refuse to take on riskier cases maintain high patient survival rates, potentially avoiding nonstandard treatments as well as possible deaths from medical interventions detected by his data trends. As a result, success rates are also artificially inflated.
Yet Agus believes that tapping data’s power will lead to significant innovation. Algorithms and artificial intelligence have been around for a while, but now we’re learning how to better collect and organize data. This decade is going to be the decade of data. “We sequenced DNA and looked into its associations during the past decade.
Here are a few examples of tech innovations in health care:
Reflections on robots: Sometimes, robots are at play. Despite having difficulty distinguishing facial expressions, children with autism during a University of Bristol study in 2017. During the same year, Dell Technologies introduced Milo, a 2-foot tall, visually expressive robot that teaches autistic kids ages 5 to 17 how to recognize signs of emotion. Milo is now being used by educational facilities in 27 states across the country.
ALS patients can now communicate again with eye-tracking glasses that use AI technology known as a brain-computer interface (BCI). Patients type with their eyes onto a monitor that voices their thoughts through computerized decoding, plus they use email, read books, and stay connected to the world.
Some types of cardiac arrhythmias, particularly atrial fibrillation, can increase the risk of heart attacks or strokes. A Stanford University study shows that AI software can identify arrhythmias from an electrocardiogram (EKG) more accurately than a human expert.
Magnetic resonance imaging (MRI) and computerized axial tomography (CT) scans will provide detailed, noninvasive views of the inner body. By using next-generation radiology tools, artificial intelligence may soon replace the need for additional tissue samples for tumor biopsies.
By the Numbers:
1 in 7,000: Number of Americans of all ages with long QT syndrome, a deadly heart disorder, who could one day be helped by Kardio Pro, an AI-powered, at-home heart monitor that detects serious and benign arrhythmias.
30%: Reduction in patient wait time before admittance, reports Johns Hopkins Hospital, after it launched a digital command center in 2016 with 22 monitors to improve patient experience, lessen risk, and streamline flow.
95.5%: Percentage of accuracy, using a special microscope, with which a deep-learning computer program identified cancer cells with precision, according to a 2016 study from UCLA published in Nature Scientific Reports.
Author Name: Lauren Paige Kennedy