It was retrospective-only, i.e. a case series on women who were known to have breast cancer, so there were zero false negatives and zero true negatives, because all patients in the study truly had cancer.
The AI system used was a ConvNet used commercially circa 2021, which is when the data for this case series were collected.
Well yes, that's the denominator for determining selectivity, which is what the headline claim is about.
Also, they need to set up their next paper:
> However, the retrospective, cancer-only design limits generalizability, highlighting the need for prospective multicenter screening trials for validation.
Does this mean that newer AI systems would perform significantly differently?
The main known way to improve performance on tasks like this is getting more data.
Please quote where I used either word.
Wouldn't this mean that AI identitied them all has having cancer?
If we're saying there was a discrepancy and we're saying that all of the patients had cancer, then it would seem that there must have been some that were identified as not having cancer by AI.
Edit: I have a problem with the way the title uses "AI" as a singular unchanging entity. It should really be "An AI system misses nearly...". There is no single AI and models are constantly improving - sometimes exponentially better.
When the treatment for a diagnosis includes radiation therapy, over-diagnosis can literally cause cancer.