When an artificial intelligence supports a radiologist in reading a mammogram, more cancers are found, earlier, and with fewer follow-up appointments. Several independent large studies published in 2024 and 2025 show this. The numbers agree so well that they put the medical discussion about AI diagnostics on a new footing.
What the Studies Show
The most extensive investigation to date comes from Germany. Researchers analyzed mammography data from 463,094 women across 119 radiology practices. The result: with AI support, 6.7 cancers were detected per 1,000 women, without AI it was 5.7. That is an increase of 17.6 percent. Extrapolated to the German mammography screening programme with around 4.2 million participants per year, this would mean more than 4,000 additional cancer diagnoses annually, most in early and treatable stages.
A South Korean study published in March 2025 in Nature Communications (AI-STREAM) produced similar numbers: a 13.8 percent higher detection rate when radiologists were supported by AI. The Swedish MASAI trial, published in Lancet Digital Health, adds another finding. Not only are more cancers detected, but so-called interval cancers, tumours that appear between two screening dates and are often more aggressive, fell by 12 percent.
Why AI Alone Is Not Enough
The research consensus is clear: AI alone is not a replacement for radiologists. Studies show that AI systems can produce higher false-positive rates than experienced doctors, depending on context. That means more unnecessary follow-up exams and avoidable stress for patients. The gain comes from the combination: AI as a second pair of eyes that reviews consistently regardless of fatigue and flags critical findings.
This is reflected in regulatory practice. The US FDA had approved around 873 AI algorithms for medical imaging by mid-2025, including roughly 115 in 2025 alone. None of these algorithms is approved as a standalone diagnostic system; all are designed as assistance systems for doctors.
Adoption Varies by Country
While Scandinavian countries have already integrated AI support into national screening programmes, other European countries are still running mostly pilot projects. The decisions now facing health systems are less technical and more about reimbursement, workflow integration and legal liability.
What This Means for Patients
Breast cancer is the most common cancer in women worldwide, with more than 2.3 million new diagnoses every year. Early detection is the single most important factor for survival and cure rates. Tumours diagnosed at stage I have a ten-year survival rate of more than 90 percent. At stage IV, it is under 30 percent. Every additional early diagnosis is therefore not abstract but concretely lifesaving.