What peer review or regulatory approval process has this been through? Seems pretty irresponsible -- there are many notorious pitfalls encountered with ML for medical imaging. You shouldn't play with people's lives.
I don't understand why this comment is downvoted. Automated screening of radiological images by means of neural net is an extensively researched topic. Ten years ago there had been predictions that such automated screening will displace the radiologists, but that clearly did not happen.
For instance, this article is silent on false positive/false negative rates of the software. There is no comparison with other research. It reads like a corporate press release promoting a product.
The quotations around "amateur" should be moved to "fought". Nonetheless, it's encouraging that this level of research can be executed at home, however the strict burden of proof required should still be maintained.
Well, he was director of R&D for a medical imaging company and worked directly with academia. So I think its appropriate to contextualise the "amateur" because his work looked much more amateur than it really was.
This is an incredibly important point. Medical research must be taken seriously and I see many problems with the processes being applied here.
(for those who care- I'm a published ml biologist who works for a pharma that develops human health products. Having worked in this area for some time, I often see people who have no real idea of how the medical establishment works, or how diagnostics are marketed/sold/regulated. Overconfidence by naive individuals can have massive negative outcomes.
Why does everyone assume this guy has zero business attempting this? If you read his credentials, he should be every bit as qualified as you to attempt this kind of work while understanding the pitfalls.
According to his CV he's been active in the field for quite some time. The default assumption that he's an idiot and going to kill people just seems too cynical here.
Grandparent - you specifically mention having noted methodology problems, would you mind sharing where in the methodology you think he's gone wrong?
I'm not going to detail challenges in medical ML - the literature can do that. But just to mention one that other people haven't that goes beyond just precision-recall: algorithms can be biased based on variance in physiology (e.g. more accurate for men), and understanding how an algorithm is biased is very important for the person interpeting the information, who should be a trained doctor.
And it doesn't matter what his credentials are, that's appeal to authority. If he thinks this should be used and trusted by people for decision making, then he should submit it to independent peer review and regulatory approval.
This is like taking your temperature at home, are you making a diagnostic yourself? Not quite. But you can know some symptoms and take action (going to the doctor) maybe with less anxiety
Thermometers are well understood, simple devices, and there are other complementary checks (e.g. does my forehead feel hot) if they fail.
This project might lead to people thinking they're in the clear and not seek appropriate medical treatment, or be overtreated due to an error. You should always talk to a qualified doctor if you're concerned about your health, and not use projects like these for decision making.
I don't see how a passive scan like this can be harmful . Ofcourse if it does show a positive one should confirm further with a biopsy or other standard confirmation diagnostic.
As for false negatives...
If you feel something is wrong you should get it checked thoroughly anyway .
you literally just said the scan has no value. THe point of a scan like this is to have absurdly low false negative and positive rates so that it's actionable. Unactionable medical diagnostics are worthless and just cost (money, fuel, time).
In an ideal world, that would be the case. However, people aren't 100% rational agents motivated by logic.
My aunt was diagnosed with Multiple Sclerosis which led her to lose her eyesight from her left eye, because she refused to get a thorough checkup by a professional, and even today whenever I tell her that I visited the doctor for an issue she has... not very good words to say (something something "you are a chicken, you're hypochodriac etc). And I'm saying this without entertaining the probability of her visiting a professional who just happened to be in a bad day, which could potentialy lead to a wrong diagnosis.
I've been hunting down my own diagnosis for symptoms everyone seems to tell me that aren't serious (nail discoloration and a 24/7 headache that feels like my arteries are pulsing, which lasts for YEARS, cold fingertips during the winter, and more).
I get what you're trying to say and I agree with the general message. However, more checkpoints to catch a potential failure are good. For example, if someone were to make a take-at-home device which scans nailfold capillaries (no reason for something like that to exist) I'd get that in a heartbeat. I'm being actively ignored by every medical professional that I have visited, and if I'm not ignored they give the minimum amount of attention, kind of like "well, it's not like you're dying so who cares?"
If you don't feel good yes you should see a doctor regardless.
Let's say you get a scan every year . If the scan is able to detect something earlier than a radiologist is able to identify i think it's worth paying attention to.
I agree with you, but what's most important is the impression that the average person who uses it will have. And I don't think most people would think of this as like "taking your temperature at home". I think most people who might upload their x-ray scans would take this a lot more seriously.
A false positive could create a lot of anxiety and emotional distress, and the patient might need to go to 2, 3, or 4 other doctors to get second opinions before they feel comfortable that they really don't have cancer.
A false negative could be even worse. A patient might think "oh, the official-looking online thing said I don't have cancer, so I don't need to wait for or consider a human radiologist's results", and not believe they need treatment.
I think it's very important that people understand that -- until more research is done -- this is still not a substitute for having a human look at your x-rays. If we could be reasonably sure that everyone (or at least a very large majority) understood this when using this tool, then I think people would have far fewer objections. But I don't think that's the case.
Having said that, I think it's safe to assume that this tool has saved lives, so it's almost certainly been a net positive for people.
A thermometer mostly tells you if you have an infection, and how close you are to your body temperature becoming a medically urgent or life-threatening situation.
In addition, these kinds of things will still miss lobular ("normal" cancers are ductal) breast cancers as they don't form lumps.
15% of the women with breast cancer are waiting for a non-invasize diagnostic imaging system that can see their cancer. The only thing that can see these is an MRI with gadolinium. And that gadolinium contrast causes issues in about 1 in 1000 women, so it can't be used as a general screen.
If your decision making process is a negative result tells you nothing and a positive result warrants immediate follow up, what’s the risk here? I’m assuming doctors recommending that women get checked for breast cancer is the primary breast cancer is tested and diagnosed which presumably wouldn’t change because someone make a website.
In this case I feel better about it because there is a natural limitation in that most people doing this will only have the scan because they are getting tested through a real clinical process. So effectively they are getting "standard of care" treatment implicitly, and all this does is accelerate their response to true positives. The worst case scenario is a false positive gives them a lot of anxiety / costs them money through trying to accelerate their real diagnosis only to find it isn't real.
