Medical AI inspects multi year mammograms to gauge cancer risk

Medical researchers at NYU Langone have designed an artificial intelligence system that estimates a patient's five year likelihood of developing breast cancer. Rather than reviewing only a recent scan, the model examines sequential 3D mammograms gathered across multiple years. By tracking subtle tissue changes over time, the system demonstrated greater predictive accuracy than traditional assessment methods.

In findings published in the American Journal of Roentgenology, the computer model outperformed both single-scan evaluations and traditional tools that rely primarily on medical history or tissue density. Interestingly, the study indicated that breast density alone might not be a reliable indicator of danger, because the software flagged high risk in some patients with lower density while finding lower risk in many with very dense breasts. As study co-investigator Dr. Laura Heacock noted, this technology could eventually help physicians tailor screening routines to individual risk profiles, potentially avoiding extra testing for lower-risk patients.

Several hurdles remain before the tool can enter widespread medical practice. The study relied on historical data from hardware produced by just one equipment maker, leaving open the question of whether the software performs equally well across different scanner models. Additionally, while automated tools can highlight hidden health trends, managing years of heavy imaging files across different clinic locations could create administrative and storage hurdles for smaller healthcare providers.