Stories
Slash Boxes
Comments

SoylentNews is people

posted by martyb on Tuesday April 18 2017, @11:13AM   Printer-friendly
from the taking-advice-to-heart dept.

http://www.sciencemag.org/news/2017/04/self-taught-artificial-intelligence-beats-doctors-predicting-heart-attacks

In the new study, Weng and his colleagues compared use of the [American College of Cardiology/American Heart Association (ACC/AHA)] guidelines with four machine-learning algorithms: random forest, logistic regression, gradient boosting, and neural networks. All four techniques analyze lots of data in order to come up with predictive tools without any human instruction. In this case, the data came from the electronic medical records of 378,256 patients in the United Kingdom. The goal was to find patterns in the records that were associated with cardiovascular events.

First, the artificial intelligence (AI) algorithms had to train themselves. They used about 78% of the data—some 295,267 records—to search for patterns and build their own internal "guidelines." They then tested themselves on the remaining records. Using record data available in 2005, they predicted which patients would have their first cardiovascular event over the next 10 years, and checked the guesses against the 2015 records. Unlike the ACC/AHA guidelines, the machine-learning methods were allowed to take into account 22 more data points, including ethnicity, arthritis, and kidney disease.

All four AI methods performed significantly better than the ACC/AHA guidelines. Using a statistic called AUC (in which a score of 1.0 signifies 100% accuracy), the ACC/AHA guidelines hit 0.728. The four new methods ranged from 0.745 to 0.764, Weng's team reports this month in PLOS ONE. The best one—neural networks—correctly predicted 7.6% more events than the ACC/AHA method, and it raised 1.6% fewer false alarms. In the test sample of about 83,000 records, that amounts to 355 additional patients whose lives could have been saved [open, DOI: 10.1371/journal.pone.0174944] [DX]. That's because prediction often leads to prevention, Weng says, through cholesterol-lowering medication or changes in diet.


Original Submission

This discussion has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
(1)
  • (Score: 2) by c0lo on Tuesday April 18 2017, @11:23AM (3 children)

    by c0lo (156) Subscriber Badge on Tuesday April 18 2017, @11:23AM (#495791) Journal

    that amounts to 355 additional patients whose lives could have been saved... That's because prediction often leads to prevention, Weng says, through cholesterol-lowering medication or changes in diet.

    Yeah, right... 355 people that could have lived a tasty life to the end of it, now condemned to a longer and bland one.

    Now, where's my crispy bacon? (As a matter of prevention,) just don't tell me I finished it all already, I'll have a heart attack.

    (grin)

    --
    https://www.youtube.com/watch?v=aoFiw2jMy-0 https://soylentnews.org/~MichaelDavidCrawford
    • (Score: 2) by takyon on Tuesday April 18 2017, @11:49AM (1 child)

      by takyon (881) <reversethis-{gro ... s} {ta} {noykat}> on Tuesday April 18 2017, @11:49AM (#495795) Journal

      I'll meet you midway. Enjoy your turkey "bacon".

      --
      [SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
      • (Score: 0) by Anonymous Coward on Tuesday April 18 2017, @04:28PM

        by Anonymous Coward on Tuesday April 18 2017, @04:28PM (#495896)

        Turkey bacon is ok. What about organic, free-range, "uncured" (which actually means cured with celery juice [healthline.com]) bacon?

        If I buy my "uncured" bacon from Whole Foods, does that give me a hipster aura with a +10 buff against cardiac disease?

    • (Score: 2) by VLM on Tuesday April 18 2017, @12:43PM

      by VLM (445) Subscriber Badge on Tuesday April 18 2017, @12:43PM (#495804)

      Probably similar to smoking. "You keep smoking those you're gonna die of lung cancer" "puff puff puff"

      It would be rather interesting to analyze the 355 borderline people. We're they gonna die of a massive one, or would just one little pill keep them going another 20 years? Maybe the neural network located a completely new unknown risk factor, perhaps they all have ingrown toenails.

  • (Score: 0) by Anonymous Coward on Tuesday April 18 2017, @12:18PM

    by Anonymous Coward on Tuesday April 18 2017, @12:18PM (#495799)

    So apples and oranges... why would they sabotage their research thus? I think the answer to this question would certainly've deserved a spot in the summary.

  • (Score: 0) by Anonymous Coward on Tuesday April 18 2017, @12:19PM

    by Anonymous Coward on Tuesday April 18 2017, @12:19PM (#495800)

    they're allowed to use more parameters.
    if they were allowed to use the same parameters, but yielded better results, that would be a big deal.

    as I understand it:
    right now, doctors have these guidelines, essentially a formula where you put in the values of different measurements, and out comes the chance of a heart attack.
    it is possible that there is a better but more complicated formula available that a neural net could stumble upon, and that would be a big deal; it's even possible the better formula can't be explicitly written down (i.e. some infinite series or smth).

    what was done was that a neural net was used, and more measurements were used as well, and the results were marginally better.
    ok, I heartily support writing a program that is installed on all doctor's computers that will automatically do this for every patient, but it's simply a matter of using a much more complicated formula; it would be unreasonable to expect human doctors to consider dozens of parameters at once.

  • (Score: 0) by Anonymous Coward on Tuesday April 18 2017, @01:18PM

    by Anonymous Coward on Tuesday April 18 2017, @01:18PM (#495819)

    See subject.

(1)