4mo ago

Nearly 100% of cancer identified by new AI, easily outperforming doctors

In what's expected to soon be commonplace, artificial intelligence is being harnessed to pick up signs of cancer more accurately than the trained human eye. This latest AI model has a near 100% success rate and serves as a clear sign of things to come.

Study - https://www.sciencedirect.com/science/article/pii/S2666990025000059

Futurology @futurology.today

Espiritdescali @futurology.today

4mo ago

Nearly 100% of cancer identified by new AI, easily outperforming doctors

newatlas.com /cancer/ai-cancer-diagnostic/

9 comments

refreshing to see a good use of ai
- Theres so many uses for it that can benefit humanity but they are trying to sell me a washing machine with it.
- There's a Lemmy community called "actually useful ai", I think you might like it, !auai@programming.dev
Judging by the upvotes to downvotes ratio, I think many people have become allergic to the word “AI,” even if it’s not a large language model and genuinely beneficial and applicable in practice. I guess that’s what happens when everything is called “AI.”
- I just don't trust it no matter how accurate it claims to be. Until I can ask the machine what it sees that informed its opinion, I won't believe it when it tells me whether I do or don't have cancer.
Pattern recognition with tolerable near-misses is where neural networks are fucking magic.
Most public issues with AI come from money-addicts strapping this sort of pretty-good-guesser directly to a radiation gun and then surprising you with it when you walk in with a broken arm.
Thank you for posting here!
The article doesn't spell out the broader context for laymen like me. Can anyone clarify some points. Are these images taken from biopsies of tissues that are already suspected of being cancerous? Is this work translatable to preventative screening in a way that I'm unfamiliar with, or is it limited to processing biopsies?
For automated tools like this, what sort of protocols are established to prevent doctors from being biased by the tool output? Does the person running the test provide their findings before they see the output of software detection?
- Seems for the tests they mixed a number of known cancerous images, from biopsied tissue, and mixed a bunch of imaged tissues, known to be cancer free. They then randomized the images and processed them with this AI, and it was able to find cancerous vs non-cancerous tissue, at a rate significantly higher than doctors.
  They are also seeing a lot of success in using similarly purposed AI for finding pathology presented in a wide variety of images, including MRI, CT, X-ray, etc.

9 comments