[♪ INTRO].

Thanks to science fiction, we all have a pretty good idea of what medical diagnosis will look like in the distant future. You’ll go to the doctor, they will sweep a handheld scanner like Star Trek’s medical tricorders over you, and then it’ll tell them what ails ya.

Cancer, broken leg, clogged arteries whatever it is, the tricorder will find it. Star Trek-style tricorders might seem pretty far off to those of us stuck here in the present,. I mean right now we don’t even have dicorders, but thanks to clever scientists and some new technology, tricorders might not be as distant as you think.

There are two big problems standing between us and tricorders:. Making them small, and making them smart. A tricorder would have to look at you from the outside and figure out what’s wrong on the inside, and today’s doctors accomplish that goal with lots of different tools, from X-rays to MRIs, depending on what they think might be wrong.

A lot of our ways of looking inside the body without cutting you open require big, bulky machines. And expensive ones. So scientists have looked for ways of making these and other medical devices smaller.

And back in 2012, a technology company called Qualcomm launched their Tricorder XPRIZE competition. Every entry was given the same set of goals: Make something less than 2.3 kilograms that can monitor a user’s general health and diagnose 10 specific conditions, like anemia, diabetes, and sleep apnea. They also had to choose at least three additional conditions from a larger list.

And the winners were announced in April of 2017. There were hundreds of entries, but first place went to DxtER, a device made by the Final Frontiers Medical Group team, and a team called the Dynamical Biomarkers Group came in second. Both of the top two teams were able to do pretty much everything they set out to do.

Their devices take a bunch of general information about the patient, like the results of a physical or a questionnaire, and also regularly monitor the patient’s vital signs, things like blood pressure or heart rate. Then, they use other extra tools that can run different tests to get a more complete picture. Each team designed their own pieces, so each has different components that check on patients in different ways.

Some of these devices can listen to breathing; others might analyze a urine sample; and others might check blood sugar or red and white blood cell counts, all without putting any needles or probes into someone’s body. All that information gets put into a computer, which analyzes it and puts it into a system that checks it against symptoms for different conditions. And then, out pops a diagnosis.

They can figure out if a patient is sick, and what they’re sick with. Neither team made something quite handheld, like a true tricorder, but they weren’t that far off: both devices are about the size of a shoebox. As impressive as these mini medical marvels are, though, they still haven’t fully solved the second big problem with building a tricorder: generalization.

Building devices that can detect problems as different as diabetes and whooping cough, all without so much as pricking your finger, is amazing. But it’s not quite enough if we want a true sci-fi future. The real goal would be a device that can detect pretty much anything that comes its way, not something designed for a selection of specific ailments.

A lot of scientists have tried attacking this problem using machine learning:. So they give a powerful computer a whole bunch of data about lots of different patients, along with the patients’ diagnoses, and then the computer gets to figure out for itself the best way of getting from data to diagnosis. Then, they see if it can correctly diagnose a whole bunch of new patients.

With enough data and enough practice, they hope that computers will eventually be able to take a bunch of standardized measurements and come up with a super-reliable diagnosis. Different algorithms have beaten doctors at finding cancer, predicting heart attacks, and even noticing what are known as “red flags,” the kind of symptoms that mean something is really seriously wrong and needs to be fixed immediately. But they were all designed to find those specific problems.

Just like the Qualcomm winners, they couldn’t do what doctors actually do: Pick the right diagnosis out of every single possibility, just based on a patient’s information. That job is much harder. You might think this would be the place for WebMD or even Google, the whole point of those things is to get you the right result, even if your search isn’t perfect.

But research has shown that tools like Google and WebMD are far worse than doctors at getting the right diagnosis when given the same amount of information. Like, not every headache means brain cancer. The more serious the problem, the worse computers tend to do.

But when doctors work with computer algorithms, each one doing what they’re good at, they get the right diagnosis almost every time. So that will probably be the future of medical tricorders, at least for a while. Companies will keep making smaller, less invasive, more creative sensors to help us figure out what’s going on inside our own bodies, hopefully at a lower cost.

But even in the far-off future of Star Trek, they still need Doctor McCoy. Thanks for watching this episode of SciShow, which was brought to you by our patrons on Patreon. If you want to help us make more videos like this, and get access to rewards like behind-the-scenes photos and blooper reels, which I believe I contributed to during the filming of this episode, you can go to patreon.com/scishow. [♪ OUTRO].