This is Google's first public claim on breast cancer screening: a research study of images from around 90K cases in which its model read screening mammograms with fewer false positives and fewer false negatives than human experts. At the time it was a lab result — but the arc that followed is what makes it worth revisiting.
The claim became a product line and then a market: Google licensed the model to medical company iCAD for deployment, an early-stage Swedish trial of 80,000 women found AI-supported screening lifted detection by 20%, and by late 2024 a study of 747,604 women showed patients paying extra for AI-enhanced mammograms were 21% more likely to have cancer detected. The 2020 paper is the origin point of that commercialization chain.
First-order effects
Radiologists reading screening mammograms face a benchmark result showing an AI model outperforming them on both error types — the immediate pressure falls on validation and workflow studies to see if the lab accuracy survives clinical conditions.
Second-order effects
Google's move from research to licensing — the later iCAD deal — turns rival imaging vendors into potential customers or competitors, and pushes hospital buyers toward tiered offerings; the eventual finding that women paid extra for AI-enhanced reads shows vendors monetizing the accuracy gap directly.
Third-order effects
If the pattern holds, screening radiology consolidates around AI-assisted workflows — consistent with the later picture where over 75% of FDA-cleared medical AI supports radiology — while Google's own diabetic retinopathy tool failing real-world deployment in Thailand stands as the caution that theoretical accuracy does not guarantee clinical integration.
The trend: Medical imaging is moving from AI as a research benchmark to AI as a paid layer inside screening, with Google's mammography result as the template case.
We are now at the point where if you don't use AI in health care, you are actively hurting your patients >> - trained on 29,000 women - outperformed six radiologists - as good as two doctors working together #AI #healthtech https://www.bbc.com/...
Today there was a lot of discussion regarding the novelty of Google's new paper on using AI for breast cancer screening in comparison to our earlier work (in the tweet below). Let me offer my take on this. https://twitter.com/... 1/N
1/5 Great paper in @nature regarding the utility for breast cancer screening ( https://nature.com/... ) but a very disappointing statement about research reproducibility: “The code used for training the models has a large number of dependencies on internal tooling, [...]
A Google-funded study published in Nature shows a model outperforming radiologists in detecting breast cancer: https://www.nature.com/... It's interesting to compare with IBM's work published last June: https://pubs.rsna.org/...
Tired of the #Google #health story on #AI and #breastcancer. Innovation = great! But again the “#tech better than human” narrative becomes fuel for building neoliberal pressure on #healthcare: AI was better than 1 doc (US standard) but not 2 (UK standard). https://www.wsj.com/...
A Google AI system caught cancers that were originally missed and reduced false-positive flags for patients who didn't actually have the disease, according to new research https://www.wsj.com/...
Artificial intelligence can help doctors do a better job of finding breast cancer on mammograms, according to a new report from researchers from Google and medical centers in the U.S. and Britain https://www.nytimes.com/...
Great summary of the Nature paper that shows promise in using Cloud ML to help diagnose breast cancer. There's still a long way to go but Google is partnering with leading medical centers to develop a real-world diagnostic tool. https://twitter.com/...
Researchers at @Google, @NorthwesternU, @imperialCR_UK & @RoyalSurrey building #AI-driven algorithms for analysis of mammograms with accuracy higher than human interpretation is a pioneering step towards the faster diagnosis & treatment of cancers.https://www.nytimes.com/ ...
.@Google #AI system beats doctors in detection tests for breast #cancer. Algorithm reduces both false positives and false negatives. #digitalhealth https://www.ft.com/...
Breast cancer screening is probably the most misunderstood of screening complexities. Only 10 of 10,000 women in their 50s will have their lives extended by annual screening mammos, but 940 will get an unnecessary biopsy. And 62 of those 10,000 will still die from breast cancer. …