Skip

"Please state the nature of the medical emergency."
February 2, 2012 7:27 AM   Subscribe

Qualcomm and the X Prize Foundation have launched a new contest: Envision and build the equivalent of Star Trek's medical tricorder, a portable health monitoring device that can remotely diagnose patients. The winner will receive $10 million.

Competition details / FAQ / press release (from last May)

More coverage from Forbes, bit-tech, neurologica blog.

Tricorders: previously on mefi. (Link in latter post is broken. See here and here for details.)
posted by zarq (85 comments total) 12 users marked this as a favorite

 
I've already envisioned it. I can take my partial prize money in small bills.
posted by DU at 7:31 AM on February 2, 2012 [2 favorites]


I've already envisioned it.

Envisioned? I've got one of these babies sitting in my mom's attic. Lights up and everything. Bitcoins only, please.
posted by Copronymus at 7:34 AM on February 2, 2012 [1 favorite]




Tell you one thing -- if scammers can remotely read credit card info now for identity theft, then a tricorder ought to be do-able.
posted by Guy_Inamonkeysuit at 7:35 AM on February 2, 2012


I was actually thinking about this kind of thing recently. Cellphones already have a lot of sensors and they're getting more all the time. I can use mine as a metal detector, and the Galaxy Nexus has a barometer (this chip)

Why not add ultrasound? I bet you could miniaturize DNA sensing chips and add other medical sensors. Possibly even low power X-rays, but there could be, er, regulatory problems with that.
Tell you one thing -- if scammers can remotely read credit card info now for identity theft, then a tricorder ought to be do-able.
They are only reading credit cards with embedded RFID chips, which apparently idiotic credit card companies have insisted on adding to cards. Why? It's not really clear but it's completely moronic.

They can't read regular credit cards.
posted by delmoi at 7:38 AM on February 2, 2012


Wait! First, I want my Dick Tracy two-way wrist video phone! Oh, I see...nevermind...
posted by VicNebulous at 7:40 AM on February 2, 2012


I'm trying to figure out a scenario here. It takes simple vital signs, like temperature and blood pressure and can diagnose 15 common diseases. Who is the potential user? It has to be someone who couldn't already take those vitals BUT would know what to do with them if they had them. Which is who exactly?

Or maybe the user won't need to understand the vitals if they just get a "you have measles!" output. But how can they check if they really have measles? Worse, what if they have something the device can't diagnose? Now they think they are healthy and that pounding brain-tumor-induced headache is just a hangover.

Someone upthread mentioned xrays and ultrasound, but how does that help anything? Without a vast amount of optical processing, a mobile device is never going to understand the output of that. Alternatively, if a human does it, mustn't I already be at the doctor's office?

I guess what I'm saying is it seems to be solving the wrong problem. It isn't going to make healthcare more widespread in poor communities, for instance, because they can't afford iPhones anyway. If a doctor carries one in, couldn't s/he already do these tests? The problem is not lack of sensors. The problem is also not a lack of access to sensors. The problem is a lack of access to *knowledge* (in the form of doctors and nurses).
posted by DU at 7:43 AM on February 2, 2012 [4 favorites]


I'm having a hard time thinking of 15 diseases where screening is of proven benefit.

Also WTF - the device has to make the diagnoses in 15-30 volunteers only. I think I can win this one using an Amiga 500 wearing a stethoscope and a white coat :

10 PRINT "Disease free!"
posted by roofus at 7:46 AM on February 2, 2012 [4 favorites]


Seems like building something like that will require more of an initial investment of 10 million bucks. Who gets the patent at the end?
posted by spicynuts at 7:46 AM on February 2, 2012 [1 favorite]


From their FAQ:

If the technology can diagnose strep throat, can it also prescribe my antibiotic?

So this is a great example because in the last 2 months my family has had something like 7 cases of strep, plus numerous ear infections. It'd be nice to be able to tell if you have strep before taking a day off of work/school, right? So just hold up your iPhone and...do what?

There are two ways to diagnose strep. There's a characteristic appearance of the throat and there's a culture test. There's no way a mobile device can do the culture test right now. And because of poor lighting, variation in humans and inexperience in the person holding the camera, I have very dim hopes for optical sensing to work.

If you limit it to temp, BP and respiratory rate, it's going to say "you are sick". But I already know that.
posted by DU at 7:48 AM on February 2, 2012


DU: "Who is the potential user?"

One of the most common questions on AskMe is "Is this serious? Should I see a doctor?" Even if the device does nothing more than tell a consumer, "Yes, this looks like it may be a problem. Go see a doctor now." it seems like it would be a step in the right direction, no?
posted by zarq at 7:48 AM on February 2, 2012 [1 favorite]


And if the device says "No, this doesn't match anything in my database"? What direction is that a step in?
posted by DU at 7:49 AM on February 2, 2012


There's already a sorta-working geiger counter app.
posted by griphus at 7:51 AM on February 2, 2012 [1 favorite]


One of the most common questions on AskMe is "Is this serious? Should I see a doctor?" Even if the device does nothing more than tell a consumer, "Yes, this looks like it may be a problem. Go see a doctor now." it seems like it would be a step in the right direction, no?

Zarq, for that I can make a machine that spits out "Go see a doctor now" every time, and have a sensitivity of 100%. But the thing it won't do well is specificity. Ie, for the cases where the machine says "you're fine, don't worry about it", how many are actually ok? A machine that makes a diagnosis without taking a careful patient history is not something I would trust.
posted by reformedjerk at 7:53 AM on February 2, 2012 [1 favorite]


DU: "And if the device says "No, this doesn't match anything in my database"? What direction is that a step in?"

Neutral. The person isn't any better or worse off than before they scanned themselves.

You didn't answer my question.
posted by zarq at 7:53 AM on February 2, 2012 [1 favorite]


Even if the device does nothing more than tell a consumer, "Yes, this looks like it may be a problem. Go see a doctor now."

I'm going to make an App that replies "GO SEE A DOCTOR" to any data you put in. Femur bone sticking out of skin? Go see a doctor. Mild feeling of general unease? Go see a doctor. Doctor's appointment at 3:30 on Tuesday? You're in the wrong app, but go see a doctor.

PATENT PENDING
posted by griphus at 7:54 AM on February 2, 2012 [11 favorites]


Someone upthread mentioned xrays and ultrasound, but how does that help anything? Without a vast amount of optical processing, a mobile device is never going to understand the output of that.

Maybe I've missed a key part of their site, but I'm not seeing anything that suggests the processing has to be performed on-board.
posted by nobody at 8:01 AM on February 2, 2012 [1 favorite]


reformedjerk: "Zarq, for that I can make a machine that spits out "Go see a doctor now" every time, and have a sensitivity of 100%. "

griphus: "I'm going to make an App that replies "GO SEE A DOCTOR" to any data you put in. "

You'd lose the contest. :)

I still believe even limited diagnostic capability would likely be an improvement over none, especially if the device were simply determining certain basic, general parameters, such as presence of fever, auscultation (including breath, heart and abdominal sounds,) blood pressure, etc., all in one device.

reformedjerk: "But the thing it won't do well is specificity. Ie, for the cases where the machine says "you're fine, don't worry about it", how many are actually ok? A machine that makes a diagnosis without taking a careful patient history is not something I would trust."

I agree.
posted by zarq at 8:03 AM on February 2, 2012


Camera to test
- bmi
- redness and burst capillaries
- spots, moles
- lopsided stance and stoops.

Ir camera to test
- temperature
- areas blood can't get to.

microphone for
- heart rate
- dodgy breathing

Chemical sensors for
- weird chemicals in breath
- rot
- cancer (the sort dogs can smell)
- blood

Seems like its a possibility, but standard tests are probably a lot better.
posted by seanyboy at 8:04 AM on February 2, 2012


Oh and ultrasound for broken bones and babies.
posted by seanyboy at 8:05 AM on February 2, 2012 [1 favorite]


You'd lose the contest. :)

Screw the contest. There's enough smartphone-owning hypochondriacs in the world to keep in furs made of gold.
posted by griphus at 8:06 AM on February 2, 2012 [3 favorites]


I'm having a hard time thinking of 15 diseases where screening is of proven benefit.

I'm a layman, but I'll take a stab (at conditions where screening or quick diagnosis in the field may be of benefit):

Cancer
Heart disease
Liver disease
Kidney failure
Gall stones
Hypertension
Diabetes
Chlamydia
Syphilis
HPV
HIV
Stroke and related neurological crises (aneurism, TIE, etc.)
Thrombosis
Stomach ulcers
Wound infection
posted by TheWhiteSkull at 8:09 AM on February 2, 2012


Yeah, anyone who figured out how to make an app to effectively screen for STDs in the privacy of one's home is made in the shade.
posted by griphus at 8:12 AM on February 2, 2012 [3 favorites]


Chemicals in breath, or better yet urine, could be good. But redness, moles and especially "lopsided stance"? You people have never programmed a computer. It's going to take years of research and testing to look for those things.

Ultrasound for broken bones and babies are a great example of the lack of sensors not being the problem. A regular person can't look at an xray and see a broken bone, let alone tell how a fetus is doing.


DU: "And if the device says "No, this doesn't match anything in my database"? What direction is that a step in?"

Neutral. The person isn't any better or worse off than before they scanned themselves.


You didn't answer mine correctly. The person is likely worse off, because now they have "evidence" that they aren't really sick.

This is a particular example of the general problem of switching from analog systems to digital ones. With analog systems, we are used to the behavior of decreasing information. The outcome gradually gets worse. If the outcome is missing, there must not have been much signal there. Digital systems don't degrade that way. They are OK OK OK...nothing. To a person used to analog systems (which is almost all natural systems humans have evolved for and most artificial ones we are used to), a "nothing" output is going to look like a low-signal input when in reality it was just out-of-band.
posted by DU at 8:13 AM on February 2, 2012 [1 favorite]


To a person used to analog systems (which is almost all natural systems humans have evolved for and most artificial ones we are used to), a "nothing" output is going to look like a low-signal input when in reality it was just out-of-band.

Think of how an analog doctor handles this case. They don't just robotically emit "SYMPTOMS DO NOT MATCH." They say things like "this isn't right, but I'm not sure what it is" or "that's a normal mole, don't worry about it". They give you that degraded, but still useful, information that actually differs from positive information that you have no disease.
posted by DU at 8:16 AM on February 2, 2012


DU: "You didn't answer mine correctly. "

No, you disagree with me. However, that doesn't make my answer incorrect.

"The person is likely worse off, because now they have "evidence" that they aren't really sick."

False equivalence. "No, this doesn't match anything in my database" ≠ "You aren't really sick." Absence of knowledge is not a diagnosis. It shouldn't be difficult for the end user to be made to understand that.

....a "nothing" output is going to look like a low-signal input when in reality it was just out-of-band.

If my thermometer doesn't give me a temperature reading, I'm not going to simply give up and assume my temperature is okay. I'll go find one that will give me the information I need. That's why doctors run tests when something seems out of the ordinary. They accumulate information until they can make a diagnosis. And if the answer from a machine is "I don't know. I don't have enough information," then logically, the next step would be to refer you to a medical professional so they can complete the process.
posted by zarq at 8:27 AM on February 2, 2012


But I don't think this is the real problem here. Even assuming you make these work, cheap and available there's still a problem: They only serve to muddy the waters for the other hundreds of diseases, conditions and ailments that the currently-underserved get. "It'll improve lives!" Well...kind of. The incidence of those 15 diseases will be lower probably. But when someone brings up universal healthcare, for instance, the response will be "just get them one of those iphone dealies". It's a stupid talking point, but pretty much all the objections to universal healthcare are stupid so that hardly prevents it.

In fact, we had an example of this exact type of thing just a few years ago, when GWB suggested poor people could sell stuff on ebay. Selling a few possessions does not make you not-poor and being able to auto-diagnose 15 diseases does not make you healthy.
posted by DU at 8:28 AM on February 2, 2012


Of course, this could just as easily go the way of Idiocracy:


Uh... This goes in your mouth. This one goes in your ear. And this one goes in your butt. Shit...hang on a second. This one...uh...this one...this one goes in your mouth.




Well, don't want to sound like a dick or nothin', but, ah...it says on your chart that you're fucked up. Uh, you talk like a fag, and your shit's all retarded.
posted by TheWhiteSkull at 8:32 AM on February 2, 2012 [4 favorites]


Just as long as nobody tries to make a cloaking device... the Treaty of Algeron won't be in effect for nearly three centuries, but still.
posted by Halloween Jack at 8:33 AM on February 2, 2012


You people have never programmed a computer. It's going to take years of research and testing to look for those things.

Whatever you do, don't tell the kinect guys this.
posted by seanyboy at 8:39 AM on February 2, 2012


DU, why do you hate cool gadgets so much? Even if you toss aside the whole "diagnosis" aspect of a device like this, you are still left with a palm-sized suite of non-invasive sensors. Before you go and say that these non-invasive sensors aren't as good as the current ones, understand that the point of bounties like this is that they drive technology forward faster than pure necessity does. Having a reliable device like this might replace a whole shelf of tools in an ambulance.

You can't even begin to do higher-level "diagnosis" until you have reliable sensors. Stepping stones, man.
posted by hellphish at 8:49 AM on February 2, 2012


I'm a Type II diabetic. I have one or those nifty portable phone thingies that lets me download apps to it.

Despite Apple demoing the ability for download Blood Glucose results directly to an iPhone over 4 years ago, the device hasn't made it to market.

I wonder why, but suspect the companies don't want it happening, even if it was technological possible. Though it might be easier if everyone had implanted chips that monitored their health. Companies would love that.
posted by Brandon Blatcher at 8:55 AM on February 2, 2012


Maybe I'm totally missing the point, but I'm just visualizing how the medical tricorder is commonly used in Trek episodes. Trained medical staff can use one single device to quickly monitor and record a wide array of vitals: Blood pressure, temperature, heart rate, various chemical and hormonal imbalances, electrolyte levels, synaptic activity, viral or bacterial infection, cellular degradation or regeneration rates, apparently check fetal development, and so on, all with one pocket-sized device. Surely that kind of capability has great benefit, especially in field medicine and triage?
posted by xedrik at 8:56 AM on February 2, 2012 [1 favorite]


"I'm trying to figure out a scenario here. It takes simple vital signs, like temperature and blood pressure and can diagnose 15 common diseases. Who is the potential user? It has to be someone who couldn't already take those vitals BUT would know what to do with them if they had them. Which is who exactly?"

When those in-drugstore clinics staffed by nurse-practitioners were first spreading, I read that something like 85% of pediatric sick visits and 60% of adult sick visits were caused by the same like 8 things for kids and 5 things for adults, all of which a nurse practitioner was competent to diagnose and treat. (I can't figure out how to google this to find an appropriate article, but they were all over the popular press at the time if someone has a good search term.)

Which is to say, probably you want it diagnosing things like ear infection and strep throat and the flu and food poisoning. (And, yes, it mentions strep throat on the FAQ page.)

If a parent could buy one for $200, or schools could buy them for $1000, that's a deal. If I could take my sick kid and tricorder him and send that data to his pediatrician, saving us a trip to the doctor's office and her a full sick-child visit, that would be an enormous savings of time and hassle (and eventually money) for everyone involved, and would leave the pediatrician more free for dealing with the serious illnesses and diagnostic complications. In poor areas the idea of a "community school" is increasingly popular, where (among other things) students and their families can seek health treatment at an in-school clinic staffed by nurses, but the nurses are limited in what they can do, and the clinics are expensive and not staffed enough hours. I can see an enormous benefit there.
posted by Eyebrows McGee at 8:59 AM on February 2, 2012 [3 favorites]


Someone upthread mentioned xrays and ultrasound, but how does that help anything? Without a vast amount of optical processing, a mobile device is never going to understand the output of that. Alternatively, if a human does it, mustn't I already be at the doctor's office?
I don't know that it would be that difficult to see a broken bone in an ultrasound, but if you can't figure it out yourself and your phone can't process it, you could email the images to a doctor, or even live-stream them. You could even outsource it to someone who's job is just to view ultrasounds all day.

Obviously end user would be someone like a school nurse, you could have people trained to use it just like we train people to use CPR.
And if the device says "No, this doesn't match anything in my database"? What direction is that a step in?
Cost savings by not having people waste finite medical resources when there's nothing wrong with them, thereby freeing up those resources to be used on people who do have problems and improve outcomes overall?

Public health officials are now telling women not to get mammograms, because so many tumors are benign, and because they give off radiation, that it would be better for the average woman not to get a mammogram (This ignores the control aspect, however - people would feel more comfortable increasing risk in order to gain control, which is why people feel safer driving rather then flying, even though flying is much safer. On the other hand, a portable scanner would give me people more control while still preventing doctors from wasting their time)

Also, let's keep in mind. Doctors screw up a lot. 195,000 people died in 2004 due to medical error. I've seen numbers around 100k when looking that up in the past.

Sure, cellphone (possibly cloud connected) that can make medical diagnoses might make errors, certainly. But so long as those errors kill less then a hundred thousand people a year in the U.S, it would be an improvement over the "just see a doctor" advice.

It's a bit like worrying that self-driving cars might accidentally kill people, while ignoring the 40k (in the US) or so people killed by human drivers each year.

There seems to be this prevalent idea of "Just go to the doctor, do whatever they say, don't think about it" But studies show that patients who do their own research get better results. Patients with consumer medical devices might also help improve results. They'd be able to scan themselves whenever they wanted, rather then only when they're at the hospital.
Chemicals in breath, or better yet urine, could be good. But redness, moles and especially "lopsided stance"? You people have never programmed a computer. It's going to take years of research and testing to look for those things.
Dude seriously? Not only have I programmed computers, I've programmed computers to do exactly those kinds of things (well not with image data, personally but other types). I'm sure it's probably beyond you but it's not impossible
posted by delmoi at 9:00 AM on February 2, 2012 [1 favorite]


Ah, the FDA just approved an iPhone add-on for checking and storing blood glucose results. Hmmm, what does my insurance company think of this?
posted by Brandon Blatcher at 9:00 AM on February 2, 2012


As a licensed EMT, I can say that a device like this would make my job easier and more efficient, but wouldn't really be revolutionary in any real sense. Our own training is little more than an algorithm that goes "vitals normal? No? THEN: GO TO HOSPITAL!"
posted by Avenger at 9:01 AM on February 2, 2012 [2 favorites]


Despite Apple demoing the ability for download Blood Glucose results directly to an iPhone over 4 years ago, the device hasn't made it to market.

I wonder why, but suspect the companies don't want it happening, even if it was technological possible. Though it might be easier if everyone had implanted chips that monitored their health. Companies would love that.
It's the FDA. Medical devices need to be FDA approved just like drugs, and the standards are really, really stringent. Even if it were possible to upload glucose data to an smartphone, selling the app without FDA approval might be illegal.

Probably a good idea overall
posted by delmoi at 9:02 AM on February 2, 2012


Yeah, anyone who figured out how to make an app to effectively screen for STDs in the privacy of one's home is made in the shade.

Yeah, but can you imagine the backlash after the first nimrod accidentally hit the facebook share button after the positive diagnosis?
posted by Mooski at 9:07 AM on February 2, 2012 [2 favorites]


Oh, and a followup:
Chemicals in breath, or better yet urine, could be good. But redness, moles and especially "lopsided stance"? You people have never programmed a computer. It's going to take years of research and testing to look for those things. -- DU
Texture Classification using Fractal Geometry for the Diagnosis of Skin Cancers

Like I said, the fact that you know a little python does not make you an expert on the statistical methods for pattern recognition. Some people actually do know how to program computers to do these things.

There are a bunch of papers on the topic but a bunch seem pay walled.
posted by delmoi at 9:09 AM on February 2, 2012 [1 favorite]


Doctor Number One: what do you call a healthy patient?

Doctor Number Two: a healthy patient is one who hasn't had a complete work-up yet.
posted by bukvich at 9:14 AM on February 2, 2012 [2 favorites]


Anyway, anyone who followed the healthcare debate knows we need to do something about exploding healthcare costs. We pay way more then other countries. Since it's a for-profit system There's no incentive in the system to save costs. In fact, the higher the costs, the better it is for the industry!

Letting end users take care of more if their problem with cheap diagnostic tools could save money, which - while it's not technically finite - medical resources are. We should be using technology to make medical resources more plentiful.

Also, like I said this would be great for school nursers. Employers could have one person on staff trained to use the device. That kind of thing.

I actually do think that in the future, most of the mental work done by doctors today will be done by computer.
posted by delmoi at 9:17 AM on February 2, 2012


Mr Blatcher:
http://www.tuaw.com/2010/09/21/iphone-connected-blood-glucose-meter-now-closer-to-reality/
posted by seanyboy at 9:19 AM on February 2, 2012


It seems like the potential use for such a device falls into two categories: making it easier for professionals to monitor bodily signs and functions, but leaving them with the task of diagnosis; and as a firstline "triage" for unskilled individuals to have some idea of their condition and where is the right place to seek help; a bit like a fancy NHS Direct. In neither case will be use it to actually diagnose a specific disease, just make the whole process of getting medical help more efficient.
posted by Jehan at 9:22 AM on February 2, 2012


Texture Classification using Fractal Geometry for the Diagnosis of Skin Cancers

Like I said, the fact that you know a little python does not make you an expert on the statistical methods for pattern recognition. Some people actually do know how to program computers to do these things.

There are a bunch of papers on the topic but a bunch seem pay walled.


Right, and the fact that you don't get screened for skin cancer by someone pointing a camera at you and computing fractal geometry means, in general, that the technique doesn't provide the specificity and robustness needed to actually be a clinically-workable technique and pass the required review processes. In other words, it's going to take years of continued research, just like DU said.

To be clear, I think working to advance the development of the sensors and algorithms is a good thing, but it's also a good thing to be realistic about the current state of the art and not conflate academic research papers with something that is fieldable.
posted by BlueDuke at 9:27 AM on February 2, 2012 [1 favorite]


"It isn't going to make healthcare more widespread in poor communities, for instance, because they can't afford iPhones anyway."

In five years even the poorest countries with be awash in iPhones. Probably better than the ones we have now. You read it here first.
posted by a_girl_irl at 9:31 AM on February 2, 2012


I have developed a relationship tricorder. Point it at your partner, and 75% of the time it will display "DTMFA", the other 25% "Check again later".
posted by BozoBurgerBonanza at 9:39 AM on February 2, 2012 [1 favorite]


A tricorder would be most useful in the same situations it's useful in Star Trek: on away teams in hostile territory.
posted by grog at 9:47 AM on February 2, 2012


I have a stone that will already do this. You hit the person in the leg, then announce, you have pain in your leg. You hit the person in the head, then announce...
posted by dances_with_sneetches at 10:02 AM on February 2, 2012 [2 favorites]


BlueDuke: " Right, and the fact that you don't get screened for skin cancer by someone pointing a camera at you and computing fractal geometry means, in general, that the technique doesn't provide the specificity and robustness needed to actually be a clinically-workable technique and pass the required review processes."

A device was approved by the FDA for dermatologists this past November called the MelaFind. The company that makes them announced last week that they have a 3 year manufacturing contract and intend to sell the devices in the US and EU.

It is basically a computer-aided screening system for melanomas, intended to provide a doctor with more information to determine which moles may need a biopsy, and help them come to a final diagnosis. According to studies, it's pretty accurate at detecting melanomas, but also has a high false-positive rate. Here's an even-handed rundown. More. When they approved it, the FDA emphasized that MelaFind is a diagnostic aid and should not be exclusively relied on.

Worth noting that metastatic melanoma is one of the deadliest cancers. Its Stage IV five year survival rate is around 15-18%. And I'd venture to say that anything that allows for earlier detection is a good thing. But while there are still questions to whether this machine will prove to be helpful in a clinical setting, it does seem promising.
posted by zarq at 10:10 AM on February 2, 2012 [2 favorites]


Background/Approval info from the FDA: 1, 2
posted by zarq at 10:13 AM on February 2, 2012 [1 favorite]


Okay, more seriously. The device is required to diagnose 15 diseases, not diagnose 15 diseases correctly. Correct diagnosis requires more than a freaking device pointed at you or some machine that is five hundred pounds and sitting in your doctor's office or any hospital. It requires differential diagnosis that rules out other possibilities. And those other possibilities are based on such things as "Do you also have pain here?" "How long has this been going on?" "What medications are you already taking? Maybe this is a side effect."

These other questions are important. To turn over to consumers' hands a readout that can tell them a possible disease without having the tests to rule out other possibilities or to query the causes is potentially counterproductive.
posted by dances_with_sneetches at 10:14 AM on February 2, 2012 [1 favorite]


And I'd venture to say that anything that allows for earlier detection is a good thing. But while there are still questions to whether this machine will prove to be helpful in a clinical setting, it does seem promising.

Yeah, I wasn't arguing that automated diagnosis (or diagnosis assistance) is impossible; my research area is, in fact, closely related. But there is a *big* step from academic research paper to clinically-relevant device/technique. The reason for this is at least two-fold. The first is just that there is a lot of fiddly engineering required to make something that is reliable in the real world as opposed to a research lab. The second is more fundamental and has to do with the fact that, depending on how one views the relative costs of false positives and negatives, a viable discrimination method might not exist at all.
posted by BlueDuke at 10:34 AM on February 2, 2012 [1 favorite]


Right, and the fact that you don't get screened for skin cancer by someone pointing a camera at you and computing fractal geometry means, in general, that the technique doesn't provide the specificity and robustness needed to actually be a clinically-workable technique and pass the required review processes. -- BlueDuke
There's a pretty serious logical error in that paragraph.

You're assuming that every technology that's not on the market doesn't work. That's clearly not true.

That's like saying because every car can't drive itself, google's self-driving car isn't safe. But we know it is safe, it's just not on the market. Even if google wanted to start selling it tomorrow, they don't even have any factories to produce them.

Obviously stuff in the lab takes years to get on the market. It's even more difficult when there's regulatory hurdles. Just because a technique works, why would you go through the trouble of getting FDA approval if you can't make any money on it? Those were university researchers, and there are no software patents in Europe.

The other problem is that you're making a factual assumption that stuff isn't on the market. What makes you think that? The paper actually mentioned a few commercial products, but I didn't go to all the websites to see how they worked. Some of them might use similar technology. (And on preview Zarq found a link for something that will soon be on the market)

Brandon mentioned hearing about a glucose meter -> iphone addon a couple years ago, but it's not on the market yet, just passing some regulatory hurdles now. That doesn't mean the technology didn't work years ago, it just meant that at that point in time it hadn't been approved yet.

For medical technology to be approved, you need more then just a working algorithm, you need to make sure the code is error free. Even if the algorithm works perfectly, With a medical device if there's a buffer overflow in the network stack, it could kill someone. (i.e. a glucose meter giving inaccurate readings due to wireless static)
To be clear, I think working to advance the development of the sensors and algorithms is a good thing, but it's also a good thing to be realistic about the current state of the art and not conflate academic research papers with something that is fieldable. -- BlueDuke
Uh that's nice but what does that have to do with the future which is what we're talking about in this thread? DU's comment seemed to imply he didn't think it was possible at all and people in this thread only thought it was because they "didn't program computers".

In five years even the poorest countries with be awash in iPhones. Probably better than the ones we have now. You read it here first. - a_girl_irl
Right, you can get an old one for free without a contract already. And who said this stuff was going to be iPhone only? the Indian government is giving out android tablets for $35. Stuff like this would be great for rural health workers in the 3rd world, obviously.

It's more likely that you'd see this stuff on android, since anyone can slap android on whatever device they make, rather then something that plugs into an iPhone using the headphone port, which is the only way to attach stuff to it currently, while you can put whatever connectors you want on an android device, and many have USB host mode.
posted by delmoi at 10:34 AM on February 2, 2012 [2 favorites]


Oh, and of the interesting things in this discussion of melanoma specifically is that there seems to be an assumption that doctors do a good job of diagnosing melenomas just by looking at them. Is that true? I was curious so I looked it up. I found this:
The diagnostic accuracy for two dermatologists each with > 10 years experience in dermatology was 80%, with sensitivity of 91% and positive predictive value of 86%. Diagnostic accuracy rates for two senior registrars (each with 3–5 years experience) and six registrars (each with 1–2 years experience) were 62% and 56%, respectively.
So a doctor with just a few years of study only does a little better then chance. I'd be willing to bet that the algorithms out there today can do better then 62%. Almost certainly better then 56%.

It's possible they might even be better then 91%.

So it's likely that computers are probably already better then humans (or at least some humans), certainly humans less then 10 years experience in terms of diagnosing melanoma.

If a product could only get 62% correct, it would never be allowed on the market.

here's another paper that found about 86%.
posted by delmoi at 10:41 AM on February 2, 2012 [1 favorite]


DU, I'm in no way trying to pick an argument here, but a lot of medicine is pattern matching, and computers are great at pattern matching (although admittedly in different ways than people). I'm speaking as someone who has develops medical apps and knows a little math.

I think you're really missing the point if you don't find the sudden ubiquity of processing, connectivity and sensors very exciting. The question isn't really what you can't do, but rather what you can. Can you diagnose something interesting recording stomach gurgles and sending the audio to a server for processing? Can you do anything interesting with the camera and that bright LED on your phone?

I can understand why you might be worried about a reductive process in the hands of a layperson replacing the judgement of a physician... but on the other hand, there are lots of situations in which this might be a great development. For example, I'm working right now on an app that's mostly of interest to physicians but will be available to the public as well. I've actually had the reaction from physicians that they're excited about being able to refer patients to the app because it may make patients feel more comfortable with a course of action or inaction (e.g. not providing antibiotics for a sore throat) if the patient can look at the same algorithm that the physician does. Physicians have certainly been learning how to deal with more informed and autonomous patients for a while, now. And then there's the awesome sci-fi sweetness of doctors offices remotely pulling data from patients without them wading into the germ soup of a doctor's office.

(I definitely wouldn't jump to the conclusion that cheaper diagnostic tools are bad news for universal healthcare. Quite the opposite, I'd think.)
posted by ~ at 10:42 AM on February 2, 2012 [1 favorite]


delmoi: " If a product could only get 62% correct, it would never be allowed on the market. "

That certainly places the statistics in the medscape analysis of melafind linked above into sharper perspective:
In a prospective multicenter study, investigators compared the diagnostic performance of MelaFind with that of a group of experienced dermatologists. This study included 1383 patients with 1632 evaluated pigmented lesions, including 127 histologically confirmed MMs (45% in situ melanomas; median Breslow thickness of invasive lesions: 0.36 mm). The MelaFind device missed only 2 melanomas (125 of 127 detected), yielding a 98.4% sensitivity. This compared favorably to the average biopsy sensitivity (78%) of the dermatologists. When factoring in borderline melanocytic lesions (atypical melanocytic proliferations, high-grade or severely dysplastic nevi), the sensitivity of MelaFind remained an impressive 98.3%. For pigmented lesions that had been referred for evaluation to rule out MM, MelaFind had an average diagnostic specificity of 9.9%, superior to the clinicians' average specificity (3.7%). Unfortunately, the device also has a high false-positive rate, with one study showing that pigmented lesions identified by MelaFind as being suspicious turned out not to be melanomas in up to 15% of cases.

posted by zarq at 10:47 AM on February 2, 2012


One of the fifteen should definitely be lupus.
posted by gracedissolved at 10:48 AM on February 2, 2012 [4 favorites]


A tricorder would be most useful in the same situations it's useful in Star Trek: on away teams in hostile territory.

I don't know- it's already pretty easy to diagnose "red shirt."
posted by TheWhiteSkull at 10:58 AM on February 2, 2012 [3 favorites]


Well isn't it better to get false positives? ... er, unless there's an integrated auto-surgery attachment.
posted by sammyo at 11:24 AM on February 2, 2012 [2 favorites]


TV isn't real.
posted by telstar at 11:25 AM on February 2, 2012


Uh that's nice but what does that have to do with the future which is what we're talking about in this thread? DU's comment seemed to imply he didn't think it was possible at all and people in this thread only thought it was because they "didn't program computers".

Actually, he said
It's going to take years of research and testing to look for those things.

To which you replied
I'm sure it's probably beyond you but it's not impossible.

Seems like it's you that's misreading his statement.

Look, I work in this area. I'm excited about the confluence of processing power and sensors. I think great things are going to happen (in health/medicine and in other areas). But it's important to be realistic about the fact that pattern recognition in the real world, when you're trying to differentiate between 'oh, shit' and 'no problem' based on just a few molecules, or pixels, or what have you, is really, really freaking hard. Plus, depending on how you choose to prioritize between false positives and false negatives, there might not ever be any hope of an approach that is a 'net win' based on your evaluation criteria.
posted by BlueDuke at 11:28 AM on February 2, 2012


They can't read regular credit cards.

Sure they can -- optically with the camera (there's a startup) and magnetically with a little thing that goes into the headphone port (Square).
posted by effugas at 11:40 AM on February 2, 2012


You know, I would actually love to compete...

but as a hobby really, I've got nothing to run with electronics experience wise. Something to keep the brain motoring on about though, on the theory at least.




Man I love thinking about this already.
posted by Slackermagee at 11:44 AM on February 2, 2012 [1 favorite]




Actually, he said
It's going to take years of research and testing to look for those things.
Seems like it's you that's misreading his statement.
It wasn't one statement, he posted seven different comments, three of which I quoted in my reply. All of them seemed to be saying "this is a bad idea/ pointless" When he said "It takes years of research and testing to look for these things" he was talking about the kinds of tests we currently do not the hypotheical device.

Here's one, example of what he was saying:
I guess what I'm saying is it seems to be solving the wrong problem. It isn't going to make healthcare more widespread in poor communities, for instance, because they can't afford iPhones anyway. If a doctor carries one in, couldn't s/he already do these tests? The problem is not lack of sensors. The problem is also not a lack of access to sensors. The problem is a lack of access to *knowledge* (in the form of doctors and nurses).
He didn't say "over the next 3-5 years" or whatever. Only one of his comments mentioned any kind of time frame.

And most of them were about how he thought it was a bad idea in general because people should just ask doctors. But doctors screw up all the time. They're not that good at detecting melanoma by eye, they make mistakes that kill hundreds of thousands of Americans a year, etc.

If the argument is just "Well, this won't be practical over the next 3 years" then I don't think anyone would disagree. But we're talking about 5, 10 years out. Maybe more.

What I don't understand is why people are saying "This isn't possible with today's technology, therefore..." well actually the point isn't clear at all. We know it can't be done today that's the whole point of the contest. It took 8 years for someone to win the first X-prize.
posted by delmoi at 12:22 PM on February 2, 2012


Finally, we'll be able to detect the Klingons amongst us.
posted by dhartung at 12:41 PM on February 2, 2012


A couple of things about medical ultrasound: It doesn't work across an air gap. It doesn't work without applying a gel conductor to the transducer. If you built an ultrasound engine and transducer that did work without gel or direct contact, you'd probably have a system that was so powerful that the FDA would never approve it, because it could cook a patient's tissue. Also, ultrasound technicians do not interpret the images they capture. Doctors do that. Whether that's turf-protection or genuine necessity, a machine is not going to make diagnoses that now require a doctor. In the original Star Trek model, only doctors and nurses used the things, which argues that expert interpretation is still required.
posted by Kirth Gerson at 12:50 PM on February 2, 2012 [1 favorite]


I'm sure you all could find 435 reasons why they are bad, but I LOVE what X Prize Foundation does trying to bring "the future" into the now. I was talking to someone the other day about how iPhones etc can damn near do this.

After my medical cluster-fuck, I've bemoaned our lack of tricorders often, usually with the phrase, "I can't wait until we have Star Trek style medicine."
posted by _paegan_ at 12:58 PM on February 2, 2012


I'm looking forward to future X-Prizes.

1. Design and build a holographic doctor with appropriate skills in medicine, sarcasm, and opera singing.
2. Construct way-cool lasers to shoot everything and stuff.
3. Neelix.
posted by ArmyOfKittens at 1:41 PM on February 2, 2012 [2 favorites]


ArmyOfKittens: " 1. Design and build a holographic doctor with appropriate skills in medicine, sarcasm, and opera singing."

"Your brain is... not on file."

2. Construct way-cool lasers to shoot everything and stuff.

"They have two settings: stun and kill... It'd be best not to confuse them."

3. Neelix.

Tuvix
posted by zarq at 1:48 PM on February 2, 2012 [1 favorite]


Yeah, but can you imagine the backlash after the first nimrod accidentally hit the facebook share button after the positive diagnosis?

Feature or bug?
posted by Mitheral at 3:00 PM on February 2, 2012


TV isn't real.

Yet.
posted by His thoughts were red thoughts at 3:20 PM on February 2, 2012


Yeah, but can you imagine the backlash after the first nimrod accidentally hit the facebook share button after the positive diagnosis?

You clearly have better FB friends than me.

Tiffany's status: HERPES BITCHES!!!!1!!!!
posted by LordSludge at 4:47 PM on February 2, 2012


I'm sorry to have to point this out, but it's actually the Emergency Medical Hologram that says "Please state the nature of the medical emergency" when it appears, not the tricorder. I spend too much time thinking about what my life would be like if such things as those holograms existed. I'd like the Emergency Go To Work For You Hologram, please.
posted by feloniousmonk at 7:13 PM on February 2, 2012 [1 favorite]


feloniousmonk: "I'm sorry to have to point this out, but it's actually the Emergency Medical Hologram that says "Please state the nature of the medical emergency" when it appears, not the tricorder. I spend too much time thinking about what my life would be like if such things as those holograms existed. I'd like the Emergency Go To Work For You Hologram, please."

I'm hoping the EMH will be their next contest. :)
posted by zarq at 7:35 PM on February 2, 2012


feloniousmonk: "I'm sorry to have to point this out, but it's actually the Emergency Medical Hologram that says "Please state the nature of the medical emergency" when it appears, not the tricorder. I spend too much time thinking about what my life would be like if such things as those holograms existed. I'd like the Emergency Go To Work For You Hologram, please."

I would also want the Emergency Personal Chef, the Emergency Theraputic Masseuse, and the Emergency House Cleaner.
posted by His thoughts were red thoughts at 8:49 PM on February 2, 2012 [1 favorite]


It's sort of holograms all the way down once you get started.

I'll admit that I spend most of this daydream time thinking to myself "all I need is a shuttlecraft with a transporter and a replicator and I'm totally set for life." What an inner life I lead.
posted by feloniousmonk at 10:15 PM on February 2, 2012 [1 favorite]


Felonious Monk, while you're correct technically, it should also be noted that the EMH is capable of summoning a fully-functional holographic tricorder whether he's in the holoemitter-equipped Voyager sickbay or in the field (using his 29th-century portable holoemitter), which implies that the holoemitter sensor suite must be able to duplicate tricorder functionality completely; therefore, the EMH (or, if you insist, the equipment that creates the EMH) encompasses the tricorder. Not to mention that having some Siri-type functionality in any modern-day tricorder (the Federation's reluctance regarding artificial intelligence notwithstanding) would be a pretty damn good idea.

I'm simultaneously embarrassed and amused by how much I still dig discussions like this
posted by Halloween Jack at 8:06 AM on February 3, 2012


Halloween Jack: "it should also be noted that the EMH is capable of summoning a fully-functional holographic tricorder "

He can? I remember a scene in sickbay during the very first episode of Voyager where the doctor asks Harry Kim for a tricorder. Kim hands the doc one off his belt, who flips it open, then closes it and hands it back indignantly while sticking out his hand. "MEDICAL tricorder!"
posted by zarq at 9:09 AM on February 3, 2012 [2 favorites]


DU: Someone upthread mentioned xrays and ultrasound, but how does that help anything? Without a vast amount of optical processing, a mobile device is never going to understand the output of that. Alternatively, if a human does it, mustn't I already be at the doctor's office?

So, in your assumptions doctors only do their work at well-equipped doctor's offices? These people politely disagree.

(And I don't recall anyone other than trained medical personnel using them in ST, anyway.)

If a doctor carries one in, couldn't s/he already do these tests? The problem is not lack of sensors. The problem is also not a lack of access to sensors.

Know how I know you live in the First World?
posted by IAmBroom at 10:58 AM on February 3, 2012


As a licensed EMT, I can say that a device like this would make my job easier and more efficient, but wouldn't really be revolutionary in any real sense. Our own training is little more than an algorithm that goes "vitals normal? No? THEN: GO TO HOSPITAL!"

Checks profile for state Avenger lives in... frustrated, hopes to never get injured there.
posted by IAmBroom at 11:02 AM on February 3, 2012


zarq: Hmm. It's one of those things that I'd swear I'd seen in an episode (at least after they brought in the mobile emitter), but I can't cite one and Memory Alpha isn't backing me up, so you win... this time.
posted by Halloween Jack at 11:06 AM on February 3, 2012


I hadn't seen anything where they actually had holographic instruments, but why not? The holodeck/hologram is pretty much the go-to plot resolver in all of the series.

Someone has told me that in the book series, around the time that the Borg are eliminated, the good Dr. somehow becomes able to move around freely without having to be in a special location. I believe this same person also mentioned that in this time frame there are entire ships crewed by holograms.
posted by feloniousmonk at 8:08 PM on February 3, 2012


If you built an ultrasound engine and transducer that did work without gel or direct contact, you'd probably have a system that was so powerful that the FDA would never approve it, because it could cook a patient's tissue. Also, ultrasound technicians do not interpret the images they capture. Doctors do that.
The point is, in the future, computers will do it. Also, cooking something with sound seems rather difficult, as people expose themselves to hundreds of watts of audio and come out uncooked (if hard of hearing) all the time.

I think you might be able to build something that might damage people's ears, though (even though the base wave would be inaudible, you can still hear ultrasound due to, like, harmonics or whatever)
posted by delmoi at 10:18 AM on February 4, 2012


« Older Top five regrets of the dying   |   If you liked it, then you... Newer »


This thread has been archived and is closed to new comments



Post