More cases of breast cancer can be detected with the help of artificial intelligence

picture:

– Compared to the current standard with two radiologists, review by the AI ​​and one radiologist led to cancer being detected in four percent more women, while the doctors’ review time was 50 percent less, says the lead researcher. Frederick StrandLecturer at Department of Oncology and PathologyKarolinska Institutet.

In terms of false positive cases, that is, the proportion of those called who were healthy, which means unnecessary suffering and unnecessary health care costs, they reduced cases by six percent for the radiologists and the AI, and by 55 percent for the AI ​​as the reviewer the only. , compared to two radiologists.

-We find that for mammography screening, the AI ​​and one radiologist were a better choice than two radiologists. Unlike a previous study from Lund University, this improvement is statistically guaranteed in ScreenTrustCAD. Although much of the initial review work is handled by artificial intelligence, there is a need for doctors to carefully evaluate and, if necessary, take samples from the women who are called, says Fredrik Strand and continues:

Our study shows that AI is ready for controlled implementation in mammography screening. It is then important to ensure that the AI ​​system is evaluated well on images taken from the same type of X-ray equipment, and to continuously monitor accuracy. In the long term, there is also the possibility that artificial intelligence will take over most mammography screening evaluations.

Since June 2023, mammography screening at Capio St. Goran Hospital has been reviewed by artificial intelligence and radiologists, freeing up time for radiologists to devote time to women with breast cancer. The study was funded by the Swedish Research Council, the Stockholm Area Cancer Foundation and the software company Lunit Inc.

Publishing

“Artificial intelligence for breast cancer detection in mammography screening: a positive experience with the interventional screen for paired readers“, Karin Dimbrower, Alessio Creppa, Eugenia Colon, Martin Eklund, Fredrik Strand, The Lancet Digital Health, online September 9, 2023.