Fig. 6: Analysis of the association between breast age gap and breast cancer.

A, B show differences between healthy and breast cancer groups for internal and external datasets based on our ensembled Mammo-AGE model. We compare the breast age gap between healthy individuals and those diagnosed with breast cancer across various internal (A) and external (B) datasets. Scatter plots depict the relationship between the breast age gap and chronological age for each group. The scatter plots show that after age bias correction, the breast age gap is independent of chronological age in both healthy and cancer groups. The results demonstrate significant differences in the breast age gap between healthy and breast cancer groups across all ages and the 40–74 age groups in these datasets. Two-sided ANCOVA tests are utilized, adjusting for the chronological age and breast density (Statistical significance: ns not significant; *: P < 0.05; **: P < 0.01; ***: P < 0.001). Box plots show the median (center line), the 25th and 75th percentiles (box), and whiskers extending to data within 1.5 times the interquartile range. C Kaplan–Meier (KM) curves for inhouse (n = 10,392) and EMBED (n = 58,983) datasets. The KM curves demonstrate that individuals with a higher breast age gap have a significantly (two-sided log-rank test) lower probability of remaining healthy over time compared to those with a lower breast age gap. Shaded areas represent the 95% confidence intervals.