How can AI contribute to breast cancer screening?

AI manages to identify 99.3% of breast cancer cases detected during screening as positive

June 14, 2024, 11:17 am

artificial intelligence (IA) demonstrates high success rates in detecting breast cancer within two years of screening mammography.. Thus, the application of artificial intelligence for screening can help evaluate low-risk mammograms, thereby reducing the workload of radiologists. And all this without compromising the quality of the screening program. This was announced during the GEICAM Annual Review of Advances in Breast Cancer (RAGMA), organized by the GEICAM Breast Cancer Research Group. Researcher Solveig Hofvind, director of breast cancer screening and head of BreastScreen at the Norwegian Cancer Registry, provided more details on the existing evidence. “In a study conducted in Norway with 661,695 digital mammography examinations performed on 242,629 women, the data set included 3,807 screen-detected breast cancers and 1,110 interval breast cancers (that is, those diagnosed between mammograms) and routine screening, which appear normal). and next mammogram). When the exams taken were halved and the AI ​​used 50 percent as the threshold for low and high scores, 99.3 percent of breast tumors detected by screening were identified as positive.

(3,781 of 3,807) and 85.2 percent of interval breast cancer cases (946 of 1,110). On the other hand, 17 percent of false positives (2,725 out of 16,040) were considered negative.

Personalization using AI for verification

With these advances, AI is expected to soon enable personalized browsing experiences, but not beforeMore research is needed to provide the necessary evidence for safe implementation.

. To personalize screening, genetics must also be taken into account. Dr Montserrat García-Closas, Professor of Epidemiology at the Institute of Cancer Research and Imperial College London, discussed at RAGMA recent very large studies that allow us to know much more accurately than before a woman’s risk of developing breast cancer. for mutations in high-risk genes: how BRCA1 and BRCA2. In addition, these studies identify additional genes with high-risk mutations, representing significant advances in risk counseling for women with hereditary cancer. This expert also discussed the polygenic risk score (PRS), which can improve the ability to predict breast cancer risk, especially when combined with traditional risk factors such as family history, hormonal and reproductive factors, and breast density on mammogram.