5 thoughts on the future potential of machine learning in medicine

As new technologies emerge and extend the capabilities of physicians, researchers and scientists, the landscape of healthcare is also bound to change. One such example is machine learning, in which a computer intakes data and continually learns and improves its understanding of a concept over time. In a recent interview with STAT, Ziad Obermeyer, MD, assistant professor at Harvard Medical School and an emergency physician at Brigham and Women's Hospital in Boston, discussed how machine learning will change medicine.

1. Machine learning is a key tool in predictive analytics. But beyond that, researchers are seeking to understand how machine learning understands the rules. Dr. Obermeyer said in the interview that in machine learning, machines aren't given rules about computing data; rather, they learn the rules after taking in data and create algorithms.

"Once the algorithm has seen a million patients and what happened to them, you can show it information about a new patient and let it predict whether he might be at imminent risk for a heart attack," Dr. Obermeyer said. "These algorithms are extremely good at telling the difference. What we need to know more is, what are the rules the machine is learning, and how did it arrive at those rules? That's sort of the next frontier of this."

2. Dr. Obermeyer said machine learning doesn't show researchers and physicians how it reaches conclusions. Such algorithms only show predictions. When asked whether physicians will be willing to accept these conclusions without knowing how the algorithm got there, Dr. Obermeyer said that's just par for the course in medicine.

"In the history of medicine, there's…a lot of things that make less sense," he said. He offered the example of using steroids for immune suppression, which began after researchers noticed this type of therapy was working and then worked backward to try to understand how and why.

3. Radiologists may see the most changes in their jobs in a couple of decades, according to Dr. Obermeyer. While technology and machine learning won't necessarily eliminate their jobs, it certainly is likely to change the nature of them.

"In 20 years, radiologists won't exist in anywhere near their current form. They might look more like cyborgs: supervising algorithms reading thousands of studies per minute and zooming in to inspect and adjudicate ambiguous cases; or they might transform into 'diagnosticiains'…that go out and have more contact with patients and integrate that into their diagnostic judgments," he said.

4. Patients are likely to see gains in machine learning in the form of personalized predictions to make diagnoses, specifically regarding the end of life, Dr. Obermeyer said. "Predicting remaining life span for people is actually one of the easiest applications of machine learning," he told STAT. "It requires a unique set of data where we have electronic records linked to information about when people died. But once we have that for enough people, you can come up with a very accurate predictor of someone's likelihood of being alive one month out."

5. Of course, challenges regarding machine learning remain, particularly around sharing and accessing data in outside databases.

"I think for all the enthusiasm for machine learning in clinical medicine to date, it hasn't been matched by a burgeoning of activity in the rigorous testing of ideas and interventions," Dr. Obermeyer said. "It's all very well and good to say you've got an algorithm that's good at predicting. Now let's actually port them over to the real world in a safe and responsible and ethical way and see what happens."

To read the full interview, click here.

More articles on health IT:

9 things to know about blockchain in healthcare
Opinion: Why scribes are better than EHRs
Cigna warns of fake job listing phishing scams

© Copyright ASC COMMUNICATIONS 2019. Interested in LINKING to or REPRINTING this content? View our policies by clicking here.

 

Top 40 Articles from the Past 6 Months