In the largest NHS study ever conducted on AI in breast cancer screening, Google's artificial intelligence system didn't just match radiologists — it exceeded them. Published in *Nature Cancer* on March 10, 2026, the research found that AI identified 25% more so-called "interval cancers" than human doctors: the dangerous, aggressive tumours that slip through standard screening and only surface later as symptoms.
The study, a collaboration between Google, Imperial College London, and multiple NHS Trusts, analysed **125,000 mammograms** and represents the most comprehensive real-world evaluation of AI in breast cancer screening ever conducted. The findings are being described as a turning point for cancer medicine.
**What Are Interval Cancers — and Why Do They Matter?**
Interval cancers are the ones that fall through the cracks. They appear between routine screening appointments — either because they were truly absent at the last scan, or because they were there but too subtle for the human eye to catch. They tend to grow faster and be more aggressive than cancers detected at routine screening.
Finding 25% more of these cancers means finding them earlier, when treatment is more effective and survival rates are higher. It means fewer women receiving a diagnosis only after a lump becomes visible or symptoms become serious.
**The Full Picture: What the Data Shows**
The *Nature Cancer* paper contains multiple interlinked studies. The headline numbers are remarkable: - **25% more interval cancers detected** by AI compared to human radiologists across 125,000 mammograms - **40% reduction in radiologist workload** in a study of 50,000 women - **8.8% increase in cancer detection rate** for first-time screenings, alongside a **39.3% reduction in recall rate** (fewer unnecessary follow-up appointments) - **More cases of invasive cancer detected** overall - **Fewer false positives** — meaning fewer women anxiously recalled for biopsies that turn out to be nothing - **Scan reading time cut by almost a third** — crucial given the UK faces a projected 39% shortage of clinical radiologists by 2029
**AI in the Arbitration Chair**
One of the most interesting elements of the research was testing AI in "arbitration" — the process used when two human radiologists disagree on a diagnosis. The study found that AI could serve effectively as an arbitrator in these cases, ultimately reducing total workload and performing comparably to human arbitrators.
**"Augmentation, Not Replacement"**
Researchers were careful to emphasise that the AI is not designed to replace human radiologists — and shouldn't. The goal is augmentation: giving doctors a powerful second opinion, flagging cases that might otherwise be overlooked, and taking on the most repetitive scanning work so specialists can focus on cases that need a human mind.
*"AI should be a tool in the radiologist's hands, not a replacement for their judgment. But a tool this powerful changes what's possible."* — Imperial College London researchers
The NHS currently screens around 2 million women per year in England alone. If AI can find more cancers while simultaneously reducing the radiologist burden, the knock-on benefits for patient outcomes — and for an overstretched healthcare system — could be enormous.
Breast cancer kills 11,000 women in the UK every year. AI that finds more of it, earlier, with fewer false alarms, is one of the most tangible ways technology has ever offered to change that number. 🩺
*Sources: Nature Cancer (March 10, 2026) · Imperial College London · Google Health Blog · Digital Health · NHS England*
