Das KI-gestützte Brustkrebs-Screening identifizierte während des Screenings mehr Frauen mit klinisch relevanten Krebserkrankungen, ohne dass die Rate falsch positiver Ergebnisse höher war
https://ecancer.org/en/news/27721-ai-supported-mammography-screening-results-in-fewer-aggressive-and-advanced-breast-cancers-finds-full-results-from-first-randomised-controlled-trial
7 Kommentare
I couldn’t tell from the article, because it just generically uses the term „AI“, which is a large umbrella that doesn’t necessarily mean they’re using an LLM (but carefully doesn’t eliminate that possibility). As worded, this could be a strategy, or a combination of strategies, that have been in use/development for decades.
Don’t get me wrong, it does sound promising. However, the subtle implication that it’s an LLM seems a little off-putting.
So why do I have to pay $75 for the Al
This is what AI should be used for. It’s been hijacked by morons for the most part.
Why just women? Men die from it too
For any of y’all that aren’t familiar with breast radiology, they’ve been using AI in that field since the 90s. This isn’t really anything groundbreaking at all
I’m a long time lab specialist who spent many years behind the scope reviewing tissue sections and peripheral smears both via traditional microscopy and digital assisted review.
I find it hard to call this „AI“ nearest neighbor and „looks like“ seems more like just building very complex Boolean logic boiling down to „when I compare this data to my model it matches a positive with an acceptable confidence value.“ For example when it comes to identification of immature blood cells, an indication of many types of bone marrow and blood cancers, we have a base library/model. When we accept or reject identifications for cells that information is refining the library.
We’ve been doing this sort of work with this type of technology for over a decade now so I guess slapping AI onto it just feels bizarre to me.
It’s not so much „AI“ as it is automated to me.
I may be completely off the mark but that’s just my personal „from the field“ experience.
The important details of the study and how it was used:
„AI was used to triage examinations to single or double reading by radiologists and for detection support.“
In other words, the AI helped sort cases into „check by two doctors“ or „check by one doctor“.
Makes sense to me as a pretty safe way to use it and improve efficiency.