4 ms·

'If the "things showing up on MRIs that currently require a biopsy" don't actually pass cost-benefit analysis on their own,' So if we can't tell if it's malign

'If the "things showing up on MRIs that currently require a biopsy" don't actually pass cost-benefit analysis on their own,' So if we can't tell if it's malignant or not, you're saying we should start cancer treatment just in case? It's not a question of "tech that's stale" or "an ossified system" - it's a question of not being able to tell all malignant and benign cells apart until you look at them under a microscope. The last thing that space needs is a bunch of uninformed tech folks going "I don't know, let the market solve it, tech is great". I really recommend getting at least a basic education on the subject.

gojomo 3y ago
If it looks bad enough that you'd risk starting cancer treatment right away (!) with no further confirmation, it's definitionally something they'll have been glad to detect early. But for most things, the advice would be: watch & maybe run some other tests, if cheap & low-risk, if and only if the preponderance of the data implies it's a net-gain. If the MRIs become cheap enough, high-resolution enough, and the ML/stats-assisted classification becomes good enough – possible with more imaging/progression data – the most common thing might be: reimage in a month or two, compare results. Do more if and only if that calcs out for maximum welfare. These are ultimately probabilistic equations, not things you can hand-wave off based on old anecdotes and mere hunches about what's net worthwhile, and what's not. You have to run the tests, run the numbers, and reach a data-driven answer. On the other hand, people who don't trust the numbers, or their own ability to dispassionately follow the numbers if more data arrives, should absolutely be free to remain in blissful non-knowing. Many people can – and many more with proper coachin should be able to – use new info only if it improves their welfare. They don't have to follow automated scripts from a less-informed era.

gojomo 3y ago

If it looks bad enough that you'd risk starting cancer treatment right away (!) with no further confirmation, it's definitionally something they'll have been glad to detect early. But for most things, the advice would be: watch & maybe run some other tests, if cheap & low-risk, if and only if the preponderance of the data implies it's a net-gain. If the MRIs become cheap enough, high-resolution enough, and the ML/stats-assisted classification becomes good enough – possible with more imaging/progression data – the most common thing might be: reimage in a month or two, compare results. Do more if and only if that calcs out for maximum welfare. These are ultimately probabilistic equations, not things you can hand-wave off based on old anecdotes and mere hunches about what's net worthwhile, and what's not. You have to run the tests, run the numbers, and reach a data-driven answer. On the other hand, people who don't trust the numbers, or their own ability to dispassionately follow the numbers if more data arrives, should absolutely be free to remain in blissful non-knowing. Many people can – and many more with proper coachin should be able to – use new info only if it improves their welfare. They don't have to follow automated scripts from a less-informed era.