In NHS studies, Google AI caught breast cancers that human readers had missed

Two studies published in Nature Cancer on March 10, 2026, reported that an artificial-intelligence system reviewing breast-cancer screening scans inside Britain's National Health Service detected cancers that human radiologists had missed, while also cutting the workload of overstretched screening programs. The research was a collaboration between Google, Imperial College London and the NHS, using anonymized mammograms from routine screening.

In the first study, which analyzed scans from about 125,000 women, the AI detected 25 percent of the so-called interval cancers, the tumors that surface between scheduled screenings after an earlier scan was read as clear. Catching these earlier matters because interval cancers are often more advanced by the time symptoms appear. The system also flagged more invasive cancers and produced fewer false positives among women being screened for the first time.

“The research was a collaboration between Google, Imperial College London and the NHS, using anonymized mammograms from routine screening.”

The second study, covering more than 50,000 women, looked at how AI could ease the strain on the workforce. NHS screening uses two radiologists to read each scan, a gold standard that is hard to staff. When the AI acted as the second reader, the researchers estimated it could reduce reading workload by roughly 40 percent while maintaining the safety of double reading, freeing specialists to focus on the hardest cases.

The teams were careful about limits. A separate feasibility analysis across 12 London sites found AI is not plug-and-play: it must be calibrated to each hospital's equipment and patient population, and clinicians sometimes overruled correct AI flags, underlining that trust is built over time. The AI is a second pair of eyes, not a replacement for doctors, and large prospective trials are still needed before routine use. Even so, finding a quarter of missed cancers earlier, in real NHS data, is the kind of concrete result that could one day mean more women are treated when their cancer is most curable.