John Markoff reports that researchers at Microsoft have demonstrated that in some cases they may be able to identify internet users who are suffering from pancreatic cancer by their search queries, even before they have received a diagnosis of the disease. The researchers looked at searches conducted on Bing that indicated someone had been diagnosed with pancreatic cancer, then worked backward, looking for earlier queries that could have shown that the Bing user was experiencing symptoms before the diagnosis. Those early searches, they believe, can be warning flags.
While five-year survival rates for pancreatic cancer are extremely low, early detection of the disease can prolong life in a very small percentage of cases. The study suggests that early screening can increase the five-year survival rate of pancreatic patients to 5 to 7 percent, from just 3 percent. (Steve Jobs died in 2011 after battling pancreatic cancer for years.) ""We find that signals about patterns of queries in search logs can predict the future appearance of queries that are highly suggestive of a diagnosis of pancreatic adenocarcinoma," the researchrs write. "We show specifically that we can identify 5 to 15 percent of cases while preserving extremely low false positive rates" of as low as 1 in 100,000."
Pancreatic cancer — the fourth leading cause of cancer death in the United States – was in many ways the ideal subject for the study because it typically produces a series of subtle symptoms, like itchy skin, weight loss, light-colored stools, patterns of back pain and a slight yellowing of the eyes and skin that often don't prompt a patient to seek medical attention. Although Microsoft has no plans to develop any products linked to the discovery, Eric Horvitz, who says he lost his best childhood friend and, soon after, a close colleague in computer science to pancreatic cancer, says the stakes are too high to delay getting the word out. "People are being diagnosed too late. We believe that these results frame a new approach to pre-screening or screening, but there's work to do to go from the feasibility study to real-world fielding."
(Score: 2) by maxwell demon on Saturday June 11 2016, @07:12AM
At best it can tell if you or someone near you has pancreatic cancer. Unless you are a hermit.
The Tao of math: The numbers you can count are not the real numbers.
(Score: 0, Offtopic) by Anonymous Coward on Saturday June 11 2016, @07:14AM
If I search for GAY pr0n, will Google know I'm GAY?
(Score: 0) by Anonymous Coward on Saturday June 11 2016, @08:02AM
It just means you're "experimenting", possibly a conservative Christian looking to research the homosexual lifestyle so you can condemn them with greater ease while nervously tapping your foot.
However, if you go looking for petite girls with big butts, the jig is up as that is a bit too exact.
Hypothetically speaking of course.
(Score: -1, Troll) by Anonymous Coward on Saturday June 11 2016, @07:26AM
This presumes every person is enough of an social exhibitionist to type their symptoms into a search, instead of keeping quiet about their personal problems. But of course anyone who doesn't post every little detail of their lives to social media simply DOES NOT EXIST, right?
(Score: 2) by maxwell demon on Saturday June 11 2016, @08:08AM
No, it only assumes that if you put those symptoms into the search engine, you actually have them.
"Finding a high level of substance X in your blood tells you if you have illness Y" also does not imply that everyone does blood tests for substance X.
The Tao of math: The numbers you can count are not the real numbers.
(Score: -1, Offtopic) by Anonymous Coward on Saturday June 11 2016, @08:03AM
Thanks that really made my day!
(Score: 2) by maxwell demon on Saturday June 11 2016, @08:09AM
Wait, you didn't already know that?
The Tao of math: The numbers you can count are not the real numbers.
(Score: 0) by Anonymous Coward on Saturday June 11 2016, @08:44AM
What I am getting from this is that if you use Bing, you probably have pancratic cancer.
Whole new significance to the "blue search screen of death". Sorry, bro, sucks to be you. If only you had switched to Linux when I suggested it to you.
(Score: 1, Touché) by Anonymous Coward on Saturday June 11 2016, @08:49AM
Linux cures cancer! You heard it here on SoylentNews!
(Score: 0) by Anonymous Coward on Saturday June 11 2016, @10:01AM
Developers, developers, *throws chair at parent* developers, developers
(Score: 2, Insightful) by Anonymous Coward on Saturday June 11 2016, @09:08AM
Did I just hear Microsoft fire the starting gun on "we're using your telemetry data to try to save your life!"??
I imagine that almost all of these big companies that suck up our data tell themselves lies to justify their actions (though I presume some may be honest about their endeavors internally). They've all tried to tell us that they're tracking us like animals to "improve your web experience". And now this.
(Score: 5, Interesting) by Squidious on Saturday June 11 2016, @03:27PM
A few months ago I used Google to dig up some information about some back problems I was having. An hour later I got a phone call from a telemarketer who used my full name. "Mr. Squidious, we understand you are having some back problems? We at Company X can assist you with that." After I explained we are on all the Do Not Call lists he claimed I had filled out an online survey, which I most definitely did not.
This was the first time I had researched back pain, and the first time I received a cold call like that regarding back pain, so believe the two events are undoubtedly directly linked. I use DuckDuckGo now for everything that I can. I only fall back to Google or Bing if DDG fails utterly.
The terrorists have won, game, set, match. They've scared the people into electing authoritarian regimes.
(Score: 0) by Anonymous Coward on Saturday June 11 2016, @05:52PM
I only fall back to Google or Bing if DDG fails utterly.
So really, you're 100% DDG now, is what you're saying...
(Score: 0) by Anonymous Coward on Saturday June 11 2016, @09:59AM
Now that we can surmise that you have pancreatic cancer, we can start advertising 'cures' to you before anyone else knows that you're ill... Talk about preemptive influencing. We'll get you to recognize the brand of the highest bidder in no time and when your doctor tells you that you need medication or whatever, you'll squeal like a little piggie saying "I want to only receive meds from $HIGHEST_ADVERTISING_BIDDER because I've seem 'em before".
Behold... the power of computing, used for advertising, not for good...
(Score: 2) by tfried on Saturday June 11 2016, @11:15AM
"We show specifically that we can identify 5 to 15 percent of cases while preserving extremely low false positive rates" of as low as 1 in 100,000."
Hm. Given that the prevalence of pancreatic cancer is somewhere on the order of 10 in 100,000, that means between .5 to 1.5 true positives to 1 false positive in 100.000. Does not actually sound overly useful for improving individual treatment. Unfortunately, still of disturbingly obvious use to insurance companies and the like...
(Score: 0) by Anonymous Coward on Saturday June 11 2016, @12:44PM
> Given that the prevalence of pancreatic cancer is somewhere on the order of 10 in 100,000, that means between .5 to 1.5 true positives to 1 false positive in 100.000.
Nope.
It may be 10 out of 100,000 in the general population, I don't really know.
What I do know is that the general population are not all doing searches for those symptoms.
(Score: 2) by tfried on Saturday June 11 2016, @01:06PM
Interesting interpretation. Indeed they don't actually tell us. The 1 in 100,000 is what? One in 100,000 of people searching for any one symptom? One in 100,000 searching for a specific symptom? One in 100,000 searching for all symptoms? Or quite simply one in 100,000 of all subjects "tested"? The latter would be the usual way of reporting false positives, and it is what I assumed. But I have not read TFA, so tell me, why I am wrong.
(Score: 2) by tfried on Saturday June 11 2016, @01:11PM
Ah, ok, reading the primary source helps: It seems to be 1 error in 100,000 positive predictions.
So I was wrong, indeed, but not in the way you were thinking, either.
(Score: 2) by tfried on Saturday June 11 2016, @01:33PM
... and on even more in-depth reading, I simply got confused by the article saying stating, initially:
we focused on maintaining low false-positive rates (FPRs; ie, one wrong prediction in 100,000 correctly identified cases)
I guess I simply do not quite grasp what meaning they do mean to assign to the term FPR, in the presentation of results, the do spell it out unambiguously, in the end:
If we fixed FPR = 0.00001, we would correctly detect 52 searchers (TPs) but would mistakenly alert 30 searchers (FPs; capture cost ratio = 1.72)
Not exactly, but very close to what I had estimated, initially.
(Score: 0) by Anonymous Coward on Saturday June 11 2016, @12:49PM
The US has HIPPA to protect medical records. But it does not apply to companies like Google and Microsoft which are not medical providers. So while this research does have promise for improving health outcomes it also represents a serious threat to privacy. Microsoft could just as easily sell this information to your insurance company so that they could raise your rates before you even go to the doctor, making it practically impossible to prove they were targeting you. Or they could sell it to potential employers doing background checks so they could chose not to hire people who might get sick. The potential for underhanded uses is practically unlimited.
(Score: 0) by Anonymous Coward on Saturday June 11 2016, @02:15PM
Surely, you must mean "No Medical^W Privacy in Web Searches"
(Score: 2, Interesting) by nitehawk214 on Saturday June 11 2016, @04:14PM
I actually work with epidemiology information technology and I can say this is most likely bunk. Google Flu Trends was touted as a way to gauge flu via search history, and all it really was able to do was echo when people were being hypochondriacs. Our company has real hospital data to back that statement up.
To search for something like a specific diagnosis is not really possible from the "patient complaint" type data that someone might punch into a search engine.
"Don't you ever miss the days when you used to be nostalgic?" -Loiosh
(Score: 3, Interesting) by nitehawk214 on Saturday June 11 2016, @04:18PM
Although, all this being said, we actually have looked in to getting anonomized search data to do statistics on, despite it being useless, because we know health departments and hospitals would be stupid enough to pay for it. The actual data crunching would not be much more difficult than our usual work.
Microsoft and Google want a LOT of money for this kind of info, even anonomized to state-level; which isn't too surprising considering it is their core business of selling people's data. It was so much money and the data was so useless that it wasn't worth the money or the time to try it.
"Don't you ever miss the days when you used to be nostalgic?" -Loiosh