AI could detect breast cancer in early stages, study reveals

AI-driven software could successfully identify breast cancer missed by screenings, according to a new study

The AI system Transpara can help radiologists identify potential breast cancers earlier and faster, as the evidence-based software is already in clinical use in over 30 countries.

The incidence of breast cancer is increasing worldwide due to environmental, diet and lifestyle changes and countries are reporting a shortage of specialist breast radiologists. In the UK every mammogram is read by two specialist radiologists.

Overall, up to 25 per cent of breast cancers are missed by screening and considered detectable in retrospect.

The new study investigated over 2,000 interval cancers which were missed at the time of screening. Transpara, however, was able to independently identify up to 37.5 per cent of these exams.

Professor Nico Karssemeijer, ScreenPoint Medical’s CEO, said: “We are fortunate to be working with leading clinicians in the field to investigate breast AI and understand its strengths and limitations. We are committed to support studies that provide clinical evidence so that we can safely introduce our technology.

“This large study confirms the potential of AI to improve early detection of subtle cancers. This is a real game changer and shows that radiologists who work with AI can improve patient care significantly,” he added.

By combining Transpara breast care with breast density, it was possible to flag up to 51 per cent of women diagnosed with cancer in the interval after a negative screening.

Professor Carla van Gils of University Medical Center, who led the DENSE trial in the Netherlands, explained that: “Adding AI to breast density measurement led to a significant improvement in determining risk of an interval cancer.

“The combination of methods can help us to pinpoint the group of breast screening participants who will benefit most from supplemental MRI screening, in terms of reducing interval cancers”, she concluded.

See more info here.