back to article We know you all want to shove AI where the sun doesn't shine. And that's exactly where it's going – detecting prostate cancer

Artificially intelligent software could help doctors treat a problem that is, quite literally, a pain in the arse: prostate cancer. A team of radiologists at the University of California, Los Angeles, built a convolutional neural network to analyse MRI scans of male nether regions and detect signs of the cancer. These types …

  1. Evil Auditor Silver badge

    Risking to sound like a miserable old git, but can we please stop calling intelligent anything that includes an artificial neural network?

    Having moaned about that, it seems a promising approach, not just for old gits.

    1. Inventor of the Marmite Laser Silver badge


      It should be called FecalNet

      Nice to hear about AI getting a finger on the problem, whatever the name is.

      1. Anonymous Coward
        Anonymous Coward

        Re: Surely

        I failed the brown finger test and am meeting the Da Vinci machine next week. I'm glad that humans were able to diagnose the wee problem.

        1. Anonymous Coward
          Anonymous Coward

          Da Vinci

          In all seriousness, been there and done that - Good luck and pratice your pelvic floor muscles NOW - you are going need them.

        2. jigr1969

          Re: Surely

          Sorry to hear about that, having been there and got the T-shirt. One word of advice, along with the pelvic floor muscle exercises, measure the length and girth of your erect member before the operation. Those figures may come (pardon the expression) handy afterwards, depending on what nerves can be spared.

          Lets just say afterwards, there's a mentality of use it or lose it and one of the instruments that help you use it, will require the measurements mentioned. It will avoid an avoid conversation at a later point in time.

        3. Korev Silver badge
          Thumb Up

          Re: Surely

          Good luck AC

          1. Inventor of the Marmite Laser Silver badge

            Re: Surely

            Wot he said ^^^ times quite a lot

    2. JetSetJim

      Indeed and, while promising, training on 400 scans with a total of 728 lesions doesn't seem like nearly enough data to train on by several orders of magnitude, so very early days for this I would have thought.

  2. Anonymous Coward
    Anonymous Coward

    No more will we hear

    Twinkle twinkle little rectum

    Get your thrills when you least expect 'em

  3. Erix

    This sure gives a whole new meaning to the term ANN! Did they consider all asspects of backpropagation?

  4. James 51

    In the UK MRI capicty is going to have to be greatly expanded to use it as a prostate screening tool. This is putting the cart before the horse and screening is best targeted to at risk populations or you risk over treatment. Lots of men die with prostrate cancer rather than because of it.

    1. NATTtrash

      And on top of that: a HCP can only diagnose something IF the (potential) patient sees them. Of course we can go on and on about all kinds of scanning and "check up" initiatives, but epidemiology just shows that many, many cases are missed, just because the patients where never seen (in time). That's why it is all very good and well that AI (groan) and MRI can do the good stuff, but never should, can, or will replace the simple GP palpation when an OTAR* comes in with a runny nose.

      Oh, and a quick heads up for you gentlemen: I know that the general attitude is that is will heal itself? Well, news flash, some stuff just don't. Don't get it why you men are always so anal about it...

      *OTAR == Old Timer At Risk

      1. jigr1969

        Also, most men seem to think that the first thing that happens is the DRE (digit rectal exam), but trust me, it isn't It's a simple blood test. As my doctor told me, no doctor likes performing that and will try to avoid it if at all possible.

        1. tiggity Silver badge

          Inde a blood test, and of most use if repeated to see if anything has changed, not just done as a one off.

          Though I had to hassle my GP for one, even though I'm of an age where they should be offering these as part and parcel of routine periodic checks for things likely to crop up at some point.

          Screening is a bit sub par for a lot of conditions in the UK (hence so much cancer gets late diagnosed in UK - relative of mine probably died from prostate cancer, diagnosed late and surgery performe but he died not long after from cancer elsewhere, could be coincidence but hospital reckoned most likely metastasis from the original PC )

  5. Anonymous Coward
    Anonymous Coward

    Trusting the MRI

    Over the last 10 years I've had numerous checks for prostate cancer. Though the MRIs were clear each time they still have done the nasty template biopsies as well. ( All thankfully negative and they've finally concluded they can stop checking). Because there's a 5-15% chance* of missing something in an MRI check.

    So there's a long way to go still before the MRI route is adequate, it appears.

    *No. I don't know why it's a range.

    1. Anonymous Coward
      Anonymous Coward

      Re: Trusting the MRI

      Same here, a routine blood test showed an elevated PSA so I had a template biopsy*, that caused AUR that in turn needed a camera threaded through the urethra to put in a catheter. That stayed in for a while and eventually had a procedure** called HoLep that scoops out the prostate. Thankfully for me as well both the biopsy and HoLep results were negative.

      * Unable to go into an MRI due to having an implant.

      ** Paid for as waiting list too long 4.5k if you're interested

  6. Conundrum1885


    Nah, it nearly killed em. Get checked out already!!

    1. Evil Scot Bronze badge

      Re: Rectum?

      Ball to that.

  7. Korev Silver badge
  8. MonkeyCee

    Seems pretty good

    So it's a pretty good system, that can help with a skills shortage by assisting trained experts get a "cheap" second opinion.

    "So, anally speaking, human experts beat a neural network in an academic study that isn't in production use."

    A human trained as a physician, who then spent a decade working doing the same task. And it was fairly close. So it's pretty good at doing a SINGLE diagnostic interpretive task. From a dataset of less than a thousand cases. Plus they had the same false positive rate, which means they make similar mistakes.

    If I'd help build something like that, I'd be proud. It's not going to replace a physician, but it's going to help them do their jobs. If that means we can all get MRI scans and not die young, that would be great :)

    1. Bronek Kozicki

      Re: Seems pretty good

      Given that only 417 scans were used for training, I'd say it is very good result.

  9. BGatez

    At about 1 in 5 wrong if it were truly intelligent it would pull it's plug. At least we know why that finger was so cold.

  10. John Smith 19 Gold badge

    Once again reminds us to show respect

    for the colorectal surgeon

  11. The Oncoming Scorn Silver badge
    Thumb Up

    Machine\Artificial Intelligence

    “Here I am, brain the size of a planet, and they ask me to do that! Call that job satisfaction, 'cause I don't.........I’ve calculated your chance of survival, but I don’t think you’ll like it.”

    With apologies to DNA.

  12. William Towle

    Did you hear the other one, in which...

    ...a doctor went to a digital rectal examination and was disappointed to find out software verification *was* involved?

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like