Anshul Kundaje argues Bayes' theorem handles scanning false positives, while an analysis blames differing patient demographics

Perhaps our civilization is so far gone that we cannot handle more information. Perhaps the right move is to gouge our own eyes out, lest confusing data lead us to harm ourselves. I pray it is not so. And if it is, I pray we may yet become wise enough to handle difficult truths.

Maybe if a calibrated doctor recommends ignoring a scan result, patients can later sue the doctor into oblivion, unjustly. Our system may be that insane. But if so can we at least recognize it’s our system that is insane, not the scanner or the doctor who did the right thing?

@BenMazer Yeah they’re not. They’re very different than… cancers which are a fundamental problem here. I’m not saying it’s easy to deal with false positives. Human psychology and liability law incentives and all the rest make it difficult. But the problem is not fundamentally test results

See NELSON trial for lung cancer

https://www.nejm.org/doi/full/10.1056/NEJMoa1911793

@BenMazer I’m hearing that bayes theorem means you can just ignore incidental findings you don’t want to follow up

See DISSRM registry for small renal masses https://clinicaltrials.gov/study/NCT02346435

@aykutuz @BenMazer For decades we used to jam needles into mens' assholes to get prostate biopsies for elevated PSA in population wide screening programs. We later found out that not only much of these biopsies did not yield cancer, but also most of those with cancer did not require treatment.

@JeffLadish Thankfully machines all come with little labels that tell you what the false positives are

@InverseMarcus @JeffLadish Yes we will know exactly what to do with each and every finding that’s true

@InverseMarcus @JeffLadish Why are you assuming I’m assuming?

Watching and waiting literally cannot improve a medical outcome by the way

@elocinationn @eryney_ok The scans aren’t the bottlenecks on the core problem (rad->path correlation), which is why my enthusiasm is so moderated. It’s correlating those scans to clinical outcomes (takes time) or to tissue (takes cutting it out of someone). So all of the breathless posting about…

@BenMazer Yep. People are difficult.

“Don’t come to us for imaging for prevention before you get ill i.e when we can’t do things any worse ” is better imho.

@joejanizek @elocinationn Joe knows more than I do on this! PhD in CS, and is a radiology resident. So I doubt you can find anyone who is better suited to touch on this topic.

@BenMazer @JeffLadish The best response in a lot of these cases will be to watch and wait, or even do nothing. Why are you assuming people won't be able to do that?

@elocinationn @eryney_ok Reading Blake's more post generously as "more frequent images are going to be immensely more valuable than current screening intervals," I just am not sure that's the case. You can get an ultra-low dose lung CT follow up currently for less radiation than a chest xray for < $100.

The point is ; as a cabal; you proved your incompetence; caused irrevocable damage to humans; because of your cognitive insufficiencies which partly stem from your cabalistic dogmatism.

We now came to a point where you guys are becoming net negative road blocks for human health of future generations.

@DrMJMrBoh @BenMazer It a cabal for sure; yes: but a cabal of dogmatic incompetence; not necessarily an evil one.

@DrMJMrBoh @BenMazer Several imperfect decisions: Like ; 40 years till 1970s

@DrMJMrBoh @BenMazer Bc it is a 14yo intelligence level distraction attempt not bc it doesn’t have a good answer.

Because it is a product of human suffering not your private assets.

I may be missing your point Joe! Maybe blake is being a little sloppy with wording for dramatic effect but in general, we don't have *that* much high volume time series biological imaging data captured at close timepoints, right? And that lung cancer CT thing you posted would seem to support the premise that more is better?