HAZEL – as we’ll call her – knew something was wrong when, in her mid-50s, she started to feel short of breath at the slightest exertion. Over the next few months, she felt increasingly achy, but several medical visits and an X-ray suggested only arthritis. More troubling symptoms appeared: a persistent cough, a sore knee and tender lungs.
Whether we have had to deal with worrying symptoms or not, at some point we have all found ourselves, like Hazel, wondering what’s happening inside our own bodies. Maybe you want to know whether that cough will become a garden-variety cold or debilitating flu, or whether your child has an ear infection. At present, the only way to find out is to see a doctor. What if there were a gadget that could offer a reliable home diagnosis?
Such a device could be in your hands sooner than you think. In January last year, the X Prize Foundation partnered with communications giant Qualcomm to launch a $10 million , to be ready in mid-2015, that previously existed only in science fiction. The contest’s organisers say they want to usher in a new era of medical technology, one that would revolutionise healthcare in the face of spiralling costs and, in the US, . But just how many of a physician’s complex duties can be turned over to technology?
Advertisement
Announcing the contest last year, Peter Diamandis of the X Prize Foundation said that the guidelines were inspired by the “medical tricorder” featured in the TV series Star Trek. Waved like a wand over the human body, the smartphone-like device was capable of diagnosing myriad ailments.
Like the tricorder, the winning device must be portable, weighing no more than 2.25 kilograms. It must be able to diagnose 12 specific medical conditions – ranging from a common ear infection to pneumonia – and monitor five vital signs (see diagram). Competing teams must also choose three from a list of 12 more ambitious “elective” conditions to detect, including melanoma, food poisoning and HIV infection.
The guidelines specify that devices that work in the least invasive way will score best with the judges – a panel of non-expert users. And unlike the box of tricks on the TV show, competition devices must make diagnoses without any help from medical professionals.
Can handheld gadgets do all this? If the enthusiastic response to the competition is anything to go by, they soon will. , over 260 teams had unofficially preregistered, reflecting the fact that many of the components needed to build such a device already exist. Sensors have become powerful, small and cheap; high-resolution touchscreens are ubiquitous in phones and tablets; and cloud computing offers powerful number-crunching capabilities and access to vast online data stores.
Smartphones can already make startlingly sensitive measurements. With the right app installed, a phone can monitor your heart rate – one of the vital signs in the contest guidelines – simply by using the camera to illuminate and count the pulse in your finger. Another handheld device, the Scout, to be released later this year by contest entrant – a NASA spin-off based in Moffett Field, California – can measure four of the five vital signs on the competition’s list simply by being held to the forehead.
Inspector gadget
Monitoring vital signs is one thing. But detecting many of the medical conditions on the list requires bodily fluids to be analysed. Some competitors have already revealed devices that promise to do this without sending samples off to the lab. The ScanaFlo, another device Scanadu plans to launch this year, allows a smartphone to analyse urine and so identify two conditions on the list: urinary tract infections and type 2 diabetes. A third device, ScanaFlu, will harness a smartphone’s camera to test saliva for the early detection of conditions like strep throat and influenza. Both devices require a small immunoassay paddle, dipped into the fluid to be tested. If proteins associated with disease are present, it changes colour, which can be analysed by an app to make a diagnosis. Walter De Brouwer, Scanadu’s CEO, says it is even possible to collect and test blood with a smartphone attachment consisting of a patch of nanoneedles that are painless to use. All this strains the definition of “non-invasive”, but two emerging technologies could make it unnecessary to involve bodily fluids at all.
Some conditions can be revealed simply by the sound of your voice. Researchers at the University of Oxford demonstrated last year that Parkinson’s disease can be detected by voice-analysis software: the same tremors, weakness and rigidity that affect the limbs of Parkinson’s patients also affect the vocal cords. In laboratory tests, the software could detect the presence or absence of the disease with 99 per cent accuracy. What’s more, another study found that this vocal impairment could be detected up to five years before clinical diagnosis. Voice analysis will go far beyond Parkinson’s, says Max Little, who heads the initiative: it could also diagnose other conditions on the X Prize lists, including sleep apnea, whooping cough and stroke.
But voice analysis pales in comparison to what we might find by taking a closer look at what we exhale. Last month, researchers at the Swiss Federal Institute of Technology in Zurich revealed that all of us have unique “breathprints” that may serve as a diagnostic tool. This field has been advancing rapidly. “Twenty years ago we knew breath contained a lot of stuff but we didn’t know what biomarkers corresponded to what diseases,” says Cristina Davis, a bioinstrumentation expert at the University of California, Davis. “We can detect things now that 10 years ago you couldn’t measure.”
Bad breath
Breath analysis can, in principle, be used to diagnose , chronic obstructive pulmonary disease (COPD), pneumonia and diabetes, all on the competition’s main list. And researchers at the Cleveland Clinic in Ohio recently showed that breath tests could detect lung cancer with 75 to 80 per cent accuracy – on a par with that of a CT scan, the standard way to make the diagnosis – even before symptoms developed. “That might be early enough that it can be found at a curable stage,” says Peter Mazzone, head of Cleveland Clinic’s research on breath analysis for lung cancer detection.
Had Hazel had been able to access to such testing at home, her condition might have been diagnosed more readily. She endured nearly three years of steadily worsening symptoms, including skin that turned purple under her husband’s touch and hurt so much the pain made her jump. Finally, she sought a second opinion. “My physician wasn’t connecting the dots,” she says. By the time her new doctor referred her to specialists, what had started as lung cancer had spread to her liver, bones and brain. She was told she might have six months to live. Hazel is adamant that a simple home test would have been invaluable for her. “If I’d had a tool that could detect cancer, I would immediately have gone for a second opinion,” she says.
But even if handheld diagnostic gadgets were available, and all the sensors did their job perfectly, detecting a chemical is not the same thing as diagnosing a disease. How will such devices avoid false positives or negatives?
False positives could cause a lot of emotional harm, says Michael Epton at Christchurch Hospital in New Zealand. When making a diagnosis based on trace chemical compounds in our breath, for instance, the risk is significant. In 2009, Epton’s team found that a chemical signature they had attributed to COPD . “You have to train sensors to make sure you are detecting a specific disease,” Epton says. “You don’t want the machine to cry wolf.”
And as Hazel’s experience illustrates, the repercussions of a false negative can be even worse. Hazel says that if the device errs, she would rather it gave a false positive that could then be corrected by a physician. “It would be better than being incorrectly told you are OK,” she says. To prevent false negatives, the competition guidelines say that devices must accurately diagnose an “absence of conditions” – in other words, if the user does not have any of the 12 conditions that devices must be able to detect, then the diagnosis must say so with a high level of certainty.
Competitors could find themselves walking a thin line indeed between false positives and negatives. HIV testing, one of the elective conditions, will be especially challenging, says Anna Mastroianni, a law professor at the University of Washington in Seattle. At the moment, HIV tests in the US, whether in a clinic or at home, must be vetted by a doctor who acts as a “learned intermediary”, protecting the test-kit maker from liability in the case of an incorrect result. By having a device which is supposed to be used independently of a doctor, the contest guidelines are effectively removing that middleman. “What kind of liability does the manufacturer have?” Mastroianni says.
Granted, if your tricorder gives you a false positive, a doctor’s visit is likely to clear up the misunderstanding. The problem is that a false negative may not prompt similar vigilance.
Uneasy tensions like these may explain why teasing conclusions from the subtleties of symptoms has traditionally been left to medical professionals. “Doctors practise, they learn, there is some finesse involved,” says Jill Smith, an emergency room nurse in Baltimore, Maryland.
Self diagnosis
Smith says a tricorder of sorts is already being used and the results are not all that promising. “It’s called Dr. Google,” she says. The problem is that people trawl the web for medical information, and then often either overestimate or underestimate the significance of their own symptoms. “They’ll say, ‘oh, I only have six out of seven signs of a heart attack; I’m fine,’ and then we don’t see them until it’s too late,” she says. Even though a tricorder would perform tests to make a diagnosis, Smith says it cannot replace a physician’s training. “Until you have a computer that can reason based on things it has learned,” she says, “I don’t think you can take the art out of medicine.”
However, artificial intelligence techniques could close that gap by the time the contest ends. In recent years, AI has made vast leaps in its ability to tease conclusions out of massive data troves. , a computer scientist at the University of Washington says the AI needed to back up a successful tricorder will probably resemble Watson, the IBM supercomputer that recently beat the human champions on the US quiz show Jeopardy!. IBM is already honing Watson’s medical knowledge and working with several US hospitals to develop a virtual nurse.
A Watson-like tricorder would access the latest medical literature to inform its diagnoses. “There is more medical knowledge than doctors can keep up with today and your tricorder will be able to access that,” he says. AI would also be able to cope with what Domingos calls “fuzzy evidence”. No single measurement is usually sufficient to predict a given disease, he says: “It’s the combination of them.” So a Bayesian network – a type of probabilistic learning AI that can determine the chances of someone having a disease based on their symptoms – will likely be included in the winning design. It will also be able to learn from experience. “It can generalise beyond symptoms it has seen and learn to diagnose similar cases, not just what it has previously seen.”
“There is more medical knowledge than doctors can keep up with – but a tricorder could access it”
Perhaps the final frontier for a tricorder would be acquiring a supportive bedside manner. “Talking to a box would be a cold relationship,” says Kim Ayscue, a former nurse who is now a professor of nursing at Lynchburg College in Virginia. “Many people, especially those with chronic, long-term diseases, want more of the psychological, social side of medicine, and that is the art of being a healthcare provider.”
Even here, AI could step in. With the right natural-language user interface, a tricorder would sound genuinely sympathetic, Domingos says. “You can imagine [Apple’s digital personal assistant] Siri having a medical side to it; ‘you seem depressed. Are you congested? Do you have a runny nose?'”
With costly diagnostic equipment and procedures fuelling a rapid rise in the cost of care, a medical tricorder cannot come soon enough for some people. , equivalent to nearly one-fifth of its economic output, up from less than one-tenth 30 years ago. “The medical system is so bloated and stuck in its ways that it will take disruptive technology like this to change the status quo,” says Catherine Brownstein, a geneticist and epidemiologist at Boston Children’s Hospital. “It could make expensive diagnostic systems obsolete.”
But as Hazel’s story suggests, we could all bear in mind that sometimes no diagnosis is bulletproof, whether made by a machine or a human. A year has passed since she was told she had six months to live. The tumours that had spread throughout her body are either shrinking or held in check by aggressive chemo and radiotherapy. She says she still gets tired easily but can now swim several times a week, and she recently travelled to San Francisco to visit friends.
Anything that could flag up diseases like hers sooner, she says, should be pursued. “I think a medical tricorder would be helpful, if only to have as a monitor if you suspect something,” she says. “I thought something was strange. I had no idea what it was, and I wasn’t getting answers from anyone else.”
Build your own Star Trek medical tricorder
Can’t wait until someone gets around to winning the $10 million prize for a working medical tricorder? Get started today with a collection of apps to turn your iPhone or Android phone into a beeping, scanning, diagnosing machine.
These are just for fun and should not replace appropriate medical care. Please don’t use these apps as serious diagnostic tools – if they were that good, we’d be $10 million richer.
First things first: let’s get . For iOS, try the . It does nothing more complicated than display factoids about the television shows, but it makes all the right noises while doing it. For Android, the more immersive has customisable buttons that you can assign to your apps, so you can launch them in true “make it so” fashion.
Now to medicalise your smartphone. Get tuned into your heart rate with (Android) or (iPhone).
Track levels of glucose, proteins, nitrites, and bilirubin in your urine with (iPhone) and check your blood pressure with the addition of a little gadget called the (iPhone).
Monitor your breathing for apnea, depth, tension and other factors with (iPhone) or measure the pulse and respiration rate of yourself and others thanks to (Android).
If you’re expecting, you can count your fetus’s kicks in-utero with (Android). Later, stress yourself out by monitoring your baby’s breathing with the (Android).
Did we miss out your favourite medical app? Let us know on or .
This article appeared in print under the headline “The doctor is in your pocket”
