News for hackers. Stuff that matters.
Re:Breast cancer detection in mammography using deep learnin... by mikece on Monday December 30, 2019 @07:15PM (#21917897)
I imagine this is just scratching the surface and before long we'll be doing full-body, 3D scans every few years and everything from cancers (all of them) to heart disease, to gastero-intestinal issues, to things even as mundane as acne and dandruff will be diagnosed by algorithms pulling on a cumulative database of images of healthy and diseased body parts. The real hope is to be able to see into the brain and pick up things like CTE and Alzheimers years before symptoms manifest.
Re:Breast cancer detection in mammography using deep learnin... by Gatsky on Monday December 30, 2019 @07:56PM (#21918128)
This is the wrong idea. Detecting pathology early and then treating it is going to be complicated and expensive, and available to only a few people for a long time. We don't actually want some kind of constant AI body surveillance (for CTE?? So an NFL player can take just enough knocks before they retire?).

What we want is prevention, not early detection. Bear in mind that prevention is a proven outcome, because some people never get Alzheimers, and some people never get cancer. Detecting cancer or Alzheimers early and then intervening to cure the condition is actually not something that is definitely possible.

Re:Breast cancer detection in mammography using deep learnin... by dwelch on Monday December 30, 2019 @08:52PM (#21918361)
While I think I know what you’re getting at, there are of course examples of “detecting pathology early and then treating it” preventing morbidity, mortality, and saving money. This is the basis of all (good) public health cancer screening practices. Think of a screening colonoscopy identifying a local pre-invasive adenocarcinoma and cutting it out (cure).
Re:Breast cancer detection in mammography using deep learnin... by Gatsky on Monday December 30, 2019 @09:06PM (#21918428)
There are specific examples for high prevalence conditions as you cite, but the original comment was not talking about that, it was talking about a broader approach.

Even taking your example, getting to the point of removing a pre-invasive adenocarcinoma is actually very complicated. You need a motivated patient, you need a healthcare team of specialist doctors or nurse practitioners, pathologists, radiologists, anaesthesiologists, nurses, an endoscopy suite, a recovery area, you need a system to follow up and track these cases etc. The full embedded cost of this undertaking is huge, I don't think it is scalable, and at the same time, will never capture everyone. Randomised studies usually just barely show that these screening approaches are better than not screening, once false positives are accounted for. Furthermore, you have chosen an example where there is a useful pre-screening test (fecal blood tests), a relatively non-invasive diagnostic test (colonoscopy) and a relatively painless intervention (removing a polyp with the scope). This is not broadly applicable. A pancreatic lesion for example is very hard to diagnose for sure, and the intervention is a massive and life changing operation.

Re:Breast cancer detection in mammography using deep learnin... by jcims on Monday December 30, 2019 @10:08PM (#21918657)
We can want what we want but we have to work with what we've got.

Cancer prevention is best, I agree, but it's not always clear that it can be perfectly prevented. So then you have to move to detection and therapeutic intervention to either eliminate the cancer or prevent it from materially impacting the health of the individual. And with cancer specifically, early detection is one of the best predictors of a good outcome. This is not always possible of course, some cancers are very difficult to detect, but some aren't. It just seems like we are making perfect the enemy of good.

Re:Breast cancer detection in mammography using deep learnin... by Gatsky on Monday December 30, 2019 @10:33PM (#21918740)
What you are saying sounds reasonable, but I still feel very strongly that prevention is the key. Consider that prostate and breast cancer are extremely common, occurring in 1 in 6 men and 1 in 7 women respectively in Western countries. As someone that treats cancer, I can tell you that management of an established cancer, even if detected early, is very complex and highly unpleasant for the patient, if not only psychologically. This is unlikely to change, except it will get more expensive.

The strongest argument is an economic one. If we can develop preventative treatments that we can just give to everyone, it is a much more justifiable expenditure. The risks and benefits are much clearer. This is in contrast to what we do now, which is apply extremely expensive and complex care to a group of people that are diagnosed with cancer. There are only a small number of people in the world that can even get this care. Consider that a preventative therapy is the only type of therapy some people in the world will ever get for their cancer. Papua New Guinea for example has zero oncology specialists in a country of 8 million people.

Re:Breast cancer detection in mammography using deep learnin... by dennis_jeeves on Tuesday December 31, 2019 @04:44AM (#21919954)
Hi Gatsky, mind contacting me? Email is on my profile. ( I could not see any contact information on your profile). Admittingly the reason I wanted to contact you is partially ideological ( rest being technical/medical).
Re:Breast cancer detection in mammography using deep learnin... by jcims on Tuesday December 31, 2019 @02:42AM (#21919574)
>I still feel very strongly that prevention is the key

In case it didn't come through I'm 100% agreement with you on this. Learning about TIL and CAR-T therapies now, but in general feel that cancer vaccines would rival any pharmaceutical on the market in terms of profitability and net decrease in global misery.

Just trying to optimize the time between now and that (IMHO) inevitable day.

Re:Breast cancer detection in mammography using deep learnin... by ufo on Monday December 30, 2019 @09:35PM (#21918527)
There are reasons why this is probably not a good idea.

1. Even the best tests have a chance that the scan will produce a false positive. If we indiscriminately test everyone for every disease under the sun, most of the findings will be false positives.

2. The benefit must outweigh the cost of the test. Both in terms of expense, and intrinsic factors such as radiation exposure.

3. Certain things that show up in a scan would never cause a problem if they were just left alone. In these cases detecting it early has little benefit, and may lead to unnecessary interventions.

4. If a cure is not available, testing for the disease may not always be appropriate, from an ethics point of view.

Re:Breast cancer detection in mammography using deep learnin... by cc81 on Monday December 30, 2019 @10:30PM (#21918732)
False positives could just be us not being able to detect that they are just false positives while more advanced methods and technologies might be able to differ.
Re:Breast cancer detection in mammography using deep learnin... by ufo on Monday December 30, 2019 @11:08PM (#21918880)
That's true. The problem is that most diseases occur in a very small percentage of the population (less than 1/100 or 1/1000), which means that unless the false positive rate is exceedingly small, testing the population as a whole will still produce a bunch of false positives. This is one of the reasons why doctors only order a test if there is a reasonable suspicion that the underlying disease is present.
Re:Breast cancer detection in mammography using deep learnin... by 3fe9a03ccd14ca5 on Monday December 30, 2019 @07:22PM (#21917934)
I think I’ll pass on being a beta tester. Sounds like there’s a lot of risk from unnecessary intervention.
Re:Breast cancer detection in mammography using deep learnin... by jcims on Monday December 30, 2019 @07:42PM (#21918066)
My wife had multiple softball sized tumors growing in her abdomen for months/years that would have easily been detected by something like this. Instead they burst and metastasized and now we're a million dollars deep into medical treatments and still nothing remotely resembling a guarantee of resolution.

You can always choose not to act on what you learn. You can't act on what you don't know.

Re:Breast cancer detection in mammography using deep learnin... by jturpin on Monday December 30, 2019 @11:04PM (#21918867)
Yeah I'm on your side here. This fear of AI medicine creating false positives that overwhelm hospitals and doctors and cause unnecessary and risky surgery is one of those ideas I see floating around comments here and there, and it doesn't seem to be backed in anything tangible. I mean maybe it's a concern, but it also seems like that if it ever happened, we'd adjust for it quickly.
Re:Breast cancer detection in mammography using deep learnin... by catalogia on Monday December 30, 2019 @08:35PM (#21918285)
> You can always choose not to act on what you learn. You can't act on what you don't know.

You can also act on what you think you know but in fact don't. False negatives kill people but so can false positives.

Re:Breast cancer detection in mammography using deep learnin... by jcims on Monday December 30, 2019 @09:37PM (#21918541)
What are the numbers though? Having unnecessary procedures done because of false positives from a screening test can absolutely kill people, but so can wearing a seatbelt. People drown and burn up all the time because they couldn't get out of a wrecked vehicle.

There is presently some set of screening tests with varying levels of sensitivity and specificity, and they aren't all appropriate for mass screening. However, if millions of people started regularly getting non-ionizing imaging done through MRI or ultrasound or infrared or whatever, we would learn a shitload about predicting maladies and likely save quite a few more people than we kill in the process.

Re:Breast cancer detection in mammography using deep learnin... by catalogia on Monday December 30, 2019 @09:52PM (#21918606)
> What are the numbers though?

I couldn't say, but I wage the medical science community / industry is cognizant of the phenomena and, not being an expert in the field myself, I trust them to handle the matter reasonably. In the specific case of breast cancer, I've heard that the cost of false positives is high enough to be a major consideration in recommended breast cancer screening schedules.

Re:Breast cancer detection in mammography using deep learnin... by jabits on Monday December 30, 2019 @07:25PM (#21917958)
Only if you act on it. Just because you know or suspect something doesn’t mean you have to do something about it.