Summary

Summary

  • The technologies described in this briefing are 5 artificial intelligence (AI) technologies for mammography. They can help to select mammography images that need further diagnostic tests to see whether detected features in the image are malignant.

  • The innovative aspects are that the technologies can provide information to support radiologists, or other qualified people, to interpret mammograms.

  • The intended use would be to support qualified people to interpret mammograms.

  • The main points from the evidence summarised in this briefing are from 6 retrospective clinical validity studies, 3 conference proceedings, 3 conference abstracts, and 1 diagnostic accuracy study. They included a total of 71,470 mammography exams. They show that AI technologies may improve performance and save time in interpreting mammograms. There are a number of ongoing trials for these technologies.

  • Key uncertainties around the evidence or technology are that few clinical validity studies have used UK datasets representative of the target population for screening. Also, there are no published prospective clinical studies.

  • The cost models for the technologies are £25,000 to £45,000 for a one-off purchase plus additional licences and ongoing updates and support. Or, a yearly subscription costs £13,400 in year 1, and £11,370 thereafter, or the cost per exam subscription is between £0.60 and £3.00 per exam (excluding VAT). Using AI technologies could reduce resource use by helping reduce the workload of staff reading mammograms.