A recent cohort study published in JAMA Network Open suggests that commercial AI tools used for cancer detection in mammograms could predict breast cancer years before its clinical detection.

The study, conducted by researchers from Norway and the U.S., examined 116,495 women aged 50 to 69 who underwent at least three consecutive rounds of biennial mammography screening.

The study evaluated Lunit's AI algorithm INSIGHT MMG (version 1.1.7.2) for its ability to estimate future cancer risk based on mammograms. Results indicated that women who later developed breast cancer had consistently higher AI scores in the affected breast up to six years before diagnosis.
The mean AI score difference between breasts for women who developed screening-detected cancer increased significantly over the three rounds, from 21.3 in the first round to 79.0 in the third. For women with interval cancers, the score differences were also elevated, though less pronounced.

These findings highlight the potential of AI systems to not only detect current cancers but also estimate future risk, offering a pathway for more personalized screening strategies. "This could lead to tailored preventive measures and earlier diagnoses," the researchers noted, although further studies are required to assess the long-term impact on patient outcomes and false-positive rates.

The study also emphasized the importance of integrating AI scores with established risk models to improve prediction accuracy, suggesting AI could streamline risk-based screening approaches in the future.

Women's Health Homepage