Lancet Digit Health

AI technology enhances accuracy of breast cancer risk assessments

October 2, 2024

In this retrospective cohort study, researchers used single-cell deep learning models to search for cellular senescence, an early indicator of cancer risk, in breast biopsy samples from 4,382 healthy female donors. AI technology was trained on cells induced to senescence by ionizing radiation (IR), replicative exhaustion, or antimycin A, atazanavir-ritonavir, and doxorubicin (AAD) exposures. Primary outcome was the estimated odds of developing breast cancer, benchmarked against 5-year Gail scores.

Key findings

Among the 86 cases of breast cancer that developed over a mean of 4.8 years, researchers found significant differences in adipose-specific IR and AAD senescence prediction scores compared with controls. Higher odds of breast cancer were associated with higher scores in the adipose IR model (odds ratio [OR], 1.71; p=0.019), while the adipose AAD model showed reduced odds (OR 0.57; p=0.013). Importantly, by combining these models with Gail scores, predictive accuracy was improved, with combined predictors showing an OR of 4.70 (p<0.0001).

Source:

Heckenbach I, et al. (2024, October). Lancet Digit Health. Deep learning assessment of senescence-associated nuclear morphologies in mammary tissue from healthy female donors to predict future risk of breast cancer: a retrospective cohort study. https://pubmed.ncbi.nlm.nih.gov/39332852/