Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Midjourney Medical

midjourney.com

Post reply on HN
Print view View original

291–300 of 927 posts

Re: Midjourney Medical

#291

Post by arcticbull »

appplication wrote: post #290
arcticbull wrote: post #273

Earlier quoted context omitted.

Don’t make me tap the sign. Bayes Theorem: https://en.wikipedia.org/wiki/Bayes'_theorem There’s a very good reason we don’t test asymptomatic people in low incidence populations. Basically all positives are false positives when you do that, no matter how accurate the test is. When you’re testing healthy randos for everything the odds of a positive being false have so many 9s it would make an SRE weep. Unless this is…

So you are certainly correct but you can also tighten up your definitions for true positives as you have more information on your false positives. There may exist additional signal as well. To your point though I think there is a difference between collecting and evaluating additional data sources and using them as diagnostic tools. I suppose I fundamentally disagree with the implication of your post that there is no…

Sure collecting more data makes sense. We agree there. If that gets you to the required degree of statistical confidence my argument is moot.

Re: Midjourney Medical

#292

Post by ajmurmann »

arcticbull wrote: post #207
SpicyLemonZest wrote: post #198

Earlier quoted context omitted.

False positive rates are extremely important in the medical system as it exists today, where most scans will come without a known baseline and doctors cannot prescribe "biweekly scans for the next 6 weeks to see what changes". If we can achieve the kind of imaging abundance they're imagining (which I don't know how to evaluate based on their short post), I think false positives become much less of an issue, at least…

False positives are important because of Bayes theorem. Even a test that’s 99% sensitive in a high incidence population can be indistinguishable from noise in a low incidence population. If it has a 1% false positive rate but the incidence is 1%, the vast majority of the positives are false. Then you have to deal with the consequences, including invasive procedures for further diagnosis. If you’re searching for tens…

As we gain more data, might we be able to find patterns in that data that we now cannot see? I'm not only thinking of these regular scans but combining it with other data sources, like maybe regular, more complete blood panels, Apple Watch data, whatever we can get our hands on. Maybe we can find data points that together have a lower false-positive rate, like lump plus increased nightly body temperature plus weight loss.

Re: Midjourney Medical

#293

Post by mrandish »

carlosdp wrote: post #165
unholiness wrote: post #119

So, on the one hand, this is interesting! Reducing radiation from CT scans is a noble cause on its own. If on top of that it could make tomography cheaper and easier, you could imagine getting earlier detection of aneurisms, fibrosis, cirrhosis, thrombosis, stenosis, even plausibly cancerous masses (along with plenty of over-detection). On the other hand, nothing here substantiates this promise. We've got a video ren…

that's not a video render of a hypothetical device, that's a real video of the real working device, fwiw

> the real working device

Could you expand on the term "working"? Do you mean like "working to slowly lower a person into water while videos of animated Figma UIs play back on a monitor?" Or do you mean some crazy kind of "working", like "the ring of devices we see are scanning the organs of the woman we see and the images appearing on the monitor are those just-captured organ scans?"

Re: Midjourney Medical

#294

Post by potatoman22 »

noobermin wrote: post #64
potatoman22 wrote: post #26

Earlier quoted context omitted.

Generative models have been used in healthcare for a while for things like drug design and data generation. Not to mention all the algorithms (and probably ML) used in generating results for MRI and CT scans. I don't think this is that crazy provided they can prove it's effective.

You can't be serious about conflating a host of technologies with fucking image generation or all things. This is the worst HN comment I've seen in months and there's been loads of competition.

I was trying to say there's precedent for using ML in this field. I don't think they even said this product is an image generation model. It's probably some ML version of this https://en.wikipedia.org/wiki/Ultrasound_computer_tomography

I also found this researcher on their staff who studies tomography https://scholar.google.com/citations?hl=en&user=idvD2yYAAAAJ

Re: Midjourney Medical

#295

Post by mrtksn »

tgsovlerkhgsel wrote: post #233

I think a lot of medical diagnosis could be solved with mass data collection if it was cheap enough. Right now, blood draws are somewhat routinely done because they provide a lot of human-interpretable indicators from a small number of values, and there is some evidence that e.g. "dogs can smell cancer" etc. (i.e. some diseases cause detectable odors). With a big enough data set of [all kinds of bio values, including…

In the rest of the world diagnostics aren’t expensive at all and medical data is centralized already (blood, MRI are almost routine for hospital visits, all data stored in govt systems).

During Covid it was useful for improving protocols.

Re: Midjourney Medical

#297

Post by krzat »

arcticbull wrote: post #273
tgsovlerkhgsel wrote: post #233

I think a lot of medical diagnosis could be solved with mass data collection if it was cheap enough. Right now, blood draws are somewhat routinely done because they provide a lot of human-interpretable indicators from a small number of values, and there is some evidence that e.g. "dogs can smell cancer" etc. (i.e. some diseases cause detectable odors). With a big enough data set of [all kinds of bio values, including…

Don’t make me tap the sign. Bayes Theorem: https://en.wikipedia.org/wiki/Bayes'_theorem There’s a very good reason we don’t test asymptomatic people in low incidence populations. Basically all positives are false positives when you do that, no matter how accurate the test is. When you’re testing healthy randos for everything the odds of a positive being false have so many 9s it would make an SRE weep. Unless this is…

If this argument was as solid as you say, then all routine checks would be pointless.

I don't know about traditional blood testing, but a permanent implant which checks HR, pressure, glucose, temperature & oxidation would be pretty useful, not necessarily to diagnose anything, but to provide data for doctor when patient has actual symptomps.

Re: Midjourney Medical

#298

Post by haldujai »

adastra22 wrote: post #262
cornstalks wrote: post #246

Earlier quoted context omitted.

> Reducing radiation from CT scans is a noble cause on its own Is it? Linear No Threshold has largely been rejected at this point. https://jnm.snmjournals.org/content/early/2024/06/21/jnumed....

We have no evidence in favor of the linear no threshold model. That is not the same as saying that we have evidence against it.

There is some evidence for hormesis - but yes no model is proven right now. LNT is the most conservative model and part of why it sticks around.

A good primer: https://pmc.ncbi.nlm.nih.gov/articles/PMC2477686/

Re: Midjourney Medical

#299

Post by twostorytower »

yalok wrote: post #158
amirhirsch wrote: post #104

There are 100M pregnant women right now. If it works for just for the vanity use of seeing your baby grow (forget the medical imaging aspect) and can be as casual and relaxing experience as they put forward, then I can see such a spa being wildly successful.

is ultrasonic scanning completely harmless for developing baby? when my wife was pregnant, I remember they wouldn't recommend too frequent ultrasonic scans...

They don’t recommend them overly frequently because it’s unnecessary, but it’s not harmful to mom or the baby in any way.

Re: Midjourney Medical

#300

Post by unholiness »

arcticbull wrote: post #271
adastra22 wrote: post #262

Earlier quoted context omitted.

We have no evidence in favor of the linear no threshold model. That is not the same as saying that we have evidence against it.

Sure but we don’t prove negatives for a reason - it’s impossible. We assume the null hypothesis.

LNT is the null hypothesis. No one disagrees a linear model fits the data very well in high doses. If you want to argue that model doesn't work in low doses, you need a model with more parameters and sufficient data to fit it. The issue is that, at these low doses we want to differentiate, we're also looking at effect sizes that are hard to separate from noise, and sampling biases that are hard to erase. There's still lively and ongoing debate.
Post reply on HN