There are about 10,000 known human diseases, yet human doctors are only able to recall a fraction of them at any given moment. As many as 40,500 patients die annually in an ICU in the U.S. as a result of misdiagnosis, according to a 2012 Johns Hopkins study. British entrepreneur Ali Parsa believes that artificial intelligence can help doctors avoid these mistakes.
Parsa is the founder and CEO of Babylon, a U.K.-based subscription health service that plans to launch an AI-based app designed to improve doctors’ hit rate. Users will report the symptoms of their illness to the app, which will check them against a database of diseases using speech recognition. After taking into account the patient’s history and circumstances, Babylon will offer an appropriate course of action. Currently in beta testing, the app is expected to be available later this year.
The concept is comparable to IBM’s Watson computer, which is currently in use by oncologists at Memorial Sloan-Kettering Cancer Center in New York. IBM’s software draws from 600,000 medical evidence reports, 1.5 million patient records and clinical trials, and two million pages of text from medical journals to help doctors develop treatment plans tailored to patients’ individual symptoms, genetics, and histories.
Babylon uses a similar network of databases, though they cover illnesses beyond cancer. The system is able to analyze “hundreds of millions of combinations of symptoms” in real time, Parsa says, taking into account individualized information on the patient’s genetics, environment, behavior, and biology.
Currently, Babylon’s 150,000 registered users book doctor’s appointments and routine tests through the online service, and they can consult with one of about 100 doctors 12 hours a day, six days a week, for a cost of £7.99 ($11.40) per month. The new app, by contrast, is expected to cost £4.99 ($7.10) per month. As well as offering patients advice on sick care, it will be able to constantly monitor information on the kidneys, liver, bones, cholesterol levels, and more, along with data collected from wearable devices that monitor sleep patterns and heart rate. It issues alerts about any areas that are “red” or “amber,” in traffic-light terms, and formulates personalized health plans to keep patients in the “green,” where they are at peak health.
Parsa says the app will also be able to predict illnesses before they occur. “For example, if your heart rate is faster than normal and your physical activity hasn’t increased, it’s a sign you’re either stressed or dehydrated or you’re fighting something,” he says. “The platform can bring this to your attention and suggest the best course of action to fight the illness before it surfaces.” The app will also remind patients to take their medication, and follow up to find out how they’re feeling.
Current regulations don’t allow the app to make formal diagnoses. As a result, it is currently restricted to recommending what course of action patients should take in the immediate term. If a young person describes flu-like symptoms, the system might recommend picking up some over-the-counter medicine at a pharmacy or, if there are complicating factors in the patient's medical history, booking an appointment with a doctor. By contrast, if someone describes more serious symptoms to the app, it may recommend going straight to the hospital, or even dialing an emergency line.
Doctors will be able to view the app’s findings via a medical portal and provide consultations by text, phone, or video chat. While Parsa believes that diagnostic regulations are unlikely to change any time soon, he is also certain that AI diagnosis is less risky than diagnosis by humans. “Machines are able to recall every known disease perfectly when examining symptoms,” he says. “And unlike human doctors, they don’t have confirmation bias.”
Clare Aitchison, a medical practitioner from Norwich, is less certain. “While it’s true that computer recall is always going to be better than that of even the best doctor, what computers can’t do is communicate with people,” she says. “People describe symptoms in very different ways depending on their personalities.” Aitchison’s argument is that a human doctor who knows her patient well is able to filter what she is told and make a diagnosis accordingly.
Babylon, which recently received $25 million in funding from investors including Demis Hassabis and Mustafa Suleyman, the founders of Google’s DeepMind project, has partnered with two hospitals in Essex, where 21,500 patients are eligible to test-drive the app. The startup claims that 10 percent of eligible patients have registered to test the app and that waiting times at both hospitals have dropped since Babylon became available to their patients for free in April 2015.
Still, the risks of using artificial intelligence to offer medical advice abound. “Either it will be too sensitive and result in increased attendance at the doctor’s, in which case there isn’t much point to it,” says Aitchison, “or it won’t be sensitive enough and will result in missed serious diagnoses.”
To help avoid this situation, Parsa says, a “large number of doctors” will be “rigorously testing our symptom checker.” Used in the right manner, principally as a support for existing health-care services, Babylon could have a transformative effect, Parsa believes. “We’ll never have enough doctors to monitor each and every person’s health in this level of detail, or have the ability to analyze mountains of data in seconds,” he says. “The only way to truly democratize health care and to solve the supply-and-demand issues so many health services face is to utilize artificial intelligence, not only to help doctors meet primary-care needs but to transform the way we think about health care and switch to a model that’s preventative as well as curative.”
The big new idea for making self-driving cars that can go anywhere
The mainstream approach to driverless cars is slow and difficult. These startups think going all-in on AI will get there faster.
Inside Charm Industrial’s big bet on corn stalks for carbon removal
The startup used plant matter and bio-oil to sequester thousands of tons of carbon. The question now is how reliable, scalable, and economical this approach will prove.
The dark secret behind those cute AI-generated animal images
Google Brain has revealed its own image-making AI, called Imagen. But don't expect to see anything that isn't wholesome.
The hype around DeepMind’s new AI model misses what’s actually cool about it
Some worry that the chatter about these tools is doing the whole field a disservice.
Get the latest updates from
MIT Technology Review
Discover special offers, top stories, upcoming events, and more.