Artificial Intelligence

Why AI is about to make some of the highest-paid doctors obsolete

According to a recent New England Journal of Medicine article, radiology and pathology are particularly susceptible to the power of machine learning.

Radiologists bring home $395,000 each year, on average. Pathologists? Roughly $260,000. In the near future, however, those numbers promise to drop to $0. Don't blame Obamacare, however, or even Trumpcare (whatever that turns out to be), but rather blame the rise of machine learning and its applicability to these two areas of medicine that are heavily focused on pattern matching, a job better done by a machine than a human.

I (no longer) see dead people

Image: iStock/diego_cervo

This is the argument put forward by Dr. Ziad Obermeyer of Harvard Medical School and Brigham and Women's Hospital and Ezekiel Emanuel, PhD, of the University of Pennsylvania, in an article for the New England Journal of Medicine, one of the medical profession's most prestigious journals. Machine learning will produce big winners and losers in healthcare, according to the authors, with radiologists and pathologists among the biggest losers.

In their view, machine learning will have a disproportionately bigger impact on three areas of medicine:

  1. Machine learning will dramatically improve the ability of health professionals to establish a prognosis
  2. Machine learning will displace much of the work of radiologists and anatomical pathologists
  3. Machine learning will improve diagnostic accuracy

SEE The 6 most exciting AI advances of 2016

Honing in on the second point, the authors call out these two specialties because they're essentially asking humans to function like machines in that they rely on pattern matching. But, given enough data, machines should be far more adept at spotting a bone fracture, for example, as they write:

Massive imaging data sets, combined with recent advances in computer vision, will drive rapid improvements in performance, and machine accuracy will soon exceed that of humans. Indeed, radiology is already partway there: Algorithms can replace a second radiologist reading mammograms and will soon exceed human accuracy.

Of course, an argument can be made that machine learning will simply augment human intuition, and there's some truth to that. The problem for radiologists and pathologists, however, is the ability of machines to crunch massive quantities of data and uncover hitherto undiscoverable insights, as radiologist Nick Bryan, M.D., PhD, has discussed:

The human visual system is remarkable and radiologists learn and remember very complex patterns and use them every day to make clinical decisions and diagnoses. Modern imaging technology, however, is creating image data sets that exceed human pattern recognition capabilities. Computers and ML technology feast on such data and are rapidly becoming capable of learning incredibly complex, multi-dimensional patterns derived from large normal and diseased populations. That data may be used to diagnose known diseases, such as Alzheimer's disease, but potentially could also define new patterns for diseases such as schizophrenia.

When will my MD be rendered obsolete?

The question is when. When will this machine learning revolution sweep through radiology and pathology in a big way?

While machine learning will take years to make a dent in some industries, for radiology and pathology the future of machine learning is years, not decades, away. As Dr. Bradley Erickson, PhD, of the Mayo Clinic, has posited: "Deep-learning algorithms could begin producing radiology reports for basic studies like mammography and chest x-rays in as soon as five years, and for most types of imaging studies over the next 20 years."

While diagnostic accuracy is expected to get a boost from machine learning, Obermeyer and Emanuel expect it to take longer than the disruption of radiology and pathology. This ability to "generate differential diagnoses, suggest high-value tests and reduce overuse of testing" will emerge slowly in part because, for many conditions, the standard for diagnosis isn't well-established. In radiology and pathology, however, diagnosis tends to be "sharp and binary" (e.g., malignant vs. benign).

Whatever the impact on doctors, however, the impact of big data and machine learning on patients promises to be huge. And, along the way, if doctors are helped to get to better diagnoses faster through the aid of machines, well, that's better for both doctor and patient.

Also see

About Matt Asay

Matt Asay is a veteran technology columnist who has written for CNET, ReadWrite, and other tech media. Asay has also held a variety of executive roles with leading mobile and big data software companies.

Editor's Picks