Detection4 min read

How AI reads mammograms

AI models trained on millions of mammograms can flag suspicious areas for review — and studies show the combination catches more cancers.

Mammography is the standard screening tool for breast cancer. AI systems are trained on large sets of annotated mammograms — normal and abnormal, with biopsy-confirmed outcomes — until they learn the visual patterns that separate benign findings from suspicious ones. Given a new image, the model highlights the regions it judges concerning and assigns a risk score.

This does not happen in isolation. In clinical practice, AI acts as a second reader: it looks at every mammogram alongside the radiologist, drawing attention to subtle lesions that could go unnoticed — especially in dense breast tissue, where cancer can hide behind overlapping tissue.

What the evidence shows

  • A Swedish trial reported around 20% more cancers detected with AI double-reading
  • A large American study reported fewer false positives with AI-supported screening
  • AI performs especially well on subtle, early tumours
  • Studies compare AI against biopsy-confirmed outcomes, not guesses

AI is a decision aid, not a diagnosis. Every AI flag is reviewed by a human radiologist before any further step.

In brief

Studies report up to ~20% more cancers detected when AI supports screening.
AI models are trained on hundreds of thousands of annotated mammograms.
Dense breast tissue is where AI adds the most value.

What you can do

Ask your centre about AI-supported screening

At your next mammogram, ask whether images are read with AI help — and what the programme's evidence shows.

Understand the limits

AI flags things; your radiologist decides. Read the performance and validation articles in this section.

Keep learning

AI in breast ultrasoundAI HubAI as a second readerAI HubFDA-cleared systemsAI Hub