It turns out that schizophrenia is not just a medical condition, it’s also an apt way to think about Eric Topol’s new book, DEEP MEDICINE: How Artificial Intelligence Can Make Healthcare Human Again. Schizophrenia, after all, is characterized by inconsistent and contradictory behaviors. Much of Topol’s book seems precisely schizophrenic — breathless excitement about the transformation(s) AI will bring to medicine, coupled with recurring disclaimers about how little beneficial impact AI has had on medical practice to date.
Not surprisingly, Topol begins his book with a compelling story about AI medical success. Not long ago, a healthy newborn boy, breastfeeding well, went home from Rady Children’s Hospital in San Diego. But on his eighth day, his mother brought him to the emergency room because he was having constant seizures. Topol provides a summary medical assessment:
There was no sign of infection. A CT scan of his brain was normal; an electroencephalogram just showed the electrical signature of unending seizures. Numerous potent drugs failed to reduce the seizures; in fact, they were getting even more pronounced. The infant’s prognosis, including both brain damage and death, was bleak.
Fortunately, a blood sample was sent to Rady’s Genomic Institute for rapid whole-genome sequencing, revealing nearly 5 million locations where the baby’s genome differed from what was most common. Machine-learning algorithms quickly determined that roughly 700,000 of these variants were rare, of which 962 were known to cause diseases. Combining that information with the boy’s phenotypic data, the AI identified one, in a gene called ALDH7A1, as the most likely culprit. The variant is extremely rare, occurring in less than 0.01 percent of the population — and, notably, causes a metabolic defect that leads to seizures.
Fortunately, its effects can be overridden by dietary supplements. As soon as changes were made to the baby’s diet, his seizures abruptly ended and he was discharged home thirty-six hours later. In follow-up, he is perfectly healthy with no sign of brain damage or developmental delay.
The key to saving the baby’s life was determining the ‘needle in a haystack’ root cause of his condition — something that would have been all but impossible without Rady’s combination of genomic sequencing and artificial intelligence. And it was AI, in particular, that honed in on the problem part of the baby’s genome and, in turn, the specific life-saving remedy. Which makes this a clear and compelling success for AI-powered medicine — and pretty much the last such story in Topol’s book. Instead, several of Topol’s assessments regarding AI and medicine sound downright gloomy:
With the challenge of massive and ever-increasing data and information for each individual, no less the corpus of medical publications, it is essential that we upgrade diagnosis from an art to a digital data-driven science. Yet, so far, we have only limited prospective clinical trials to suggest this will ultimately be possible.
Similarly, here is Topol talking about the image recognition capabilities of deep neural networks:
All too commonly we ascribe the capability of machines to “read” scans or slides, when they really can’t read. Machines’ lack of understanding cannot be emphasized enough. Recognition is not understanding; there is zero context . . . A great example is the machine interpretation of “a man riding a horse down the street,” which actually is a man on a horse sitting high on a statue going nowhere. That symbolizes the plateau we’re at for image recognition.
Despite these sorts of observations, Topol remains eager for the prospect of AI-transformed medicine because he sees so clearly the downsides of current medical practice. He points out a number of specific issues, for example, that “misdiagnosis in the United States is disconcertingly common,” likely totaling approximately 12 million significant misdiagnoses a year. He also tells us that up to one-third of medical operations performed are unnecessary.
Both of which are connected, no doubt, to the fact that the average length of a clinic visit in the United States for a new patient is just twelve minutes, and for an established patient is shorter still — a mere seven minutes. Which connects, in turn, to the fact that nearly half of U.S. doctors today have symptoms of burnout, and there are hundreds of physician suicides each year. Topol concludes: “This is where we are today: patients exist in a world of insufficient data, insufficient time, insufficient context, and insufficient presence. Or, as I say, a world of shallow medicine.”
Topol clearly hopes, however, that AI will lead us into a more humane version of medicine in which new-found efficiencies are channeled into more time and attention devoted to patients. Which is why he tells us that:
What’s wrong in healthcare today is that it’s missing care . . . The greatest opportunity offered by AI is not reducing errors or workloads, or even curing cancer: it is the opportunity to restore the precious and time-honored connection and trust — the human touch — between patients and doctors.
Toward that end, Topol tells us there are a couple of specific medical fields in which AI is already having success. One of these is entirely unsurprising — the development of new pharmaceuticals, a task especially well matched to the heroic computing capabilities of AI. But the other — mental health — is quite surprising. It turns out that many humans prefer talking to a machine rather than another human when it comes to their deep, dark secrets.
By every measure, participants were willing to disclose much more when they thought they were communicating with a virtual human rather than a real one. A couple of the participants who interacted with the virtual human conveyed this well: “This is way better than talking to a person. I don’t really feel comfortable talking about personal stuff to other people.” And “A human being would be judgmental. I shared a lot of personal things, and it was because of that.”
Finally, though he acknowledges the many obstacles, Topol remains hopeful that personal AI medical assistants will eventually transform our lives. He even imagines what such interactions will be like:
“Bob, I’ve noticed that your resting heart rate and blood pressure have been climbing over the past ten days. Could you pull out your smartphone retina imaging app and take a picture?”
“OK, Rachel, here it is.”
“Bob, your retina doesn’t show any sign of your blood pressure being out of control. So that’s good. Have you been having any chest tightness?”
“With your genomic risk profile for heart disease, I just want to be sure that isn’t what is going on.”
“Thanks, Rachel. I had some peculiar sensation in my jaw the last time I was on the treadmill, but it went away after a few minutes.”
“Bob, that could be angina. I think an exercise stress test would help sort it out. I’ve looked at your schedule for next week and provisionally set this up with Dr. Jones on Tuesday afternoon at 4 P.M. on your way home from work if that’s OK.”
“Just remember to bring your running shoes and workout clothes. I’ll remind you.”
Hmmm, clearly Siri and Alexa have a long way to go.
DEEP MEDICINE was reviewed by Tim Weinhold. Read more book reviews here.