In a study reported in the Journal of Clinical Oncology, Vachon et al found that the combination of an artificial intelligence (AI) imaging algorithm, together with measures of breast density on mammography screening, were capable of long-term prediction of risk of invasive breast cancer.
As stated by the investigators, “AI algorithms improve breast cancer detection on mammography, but their contribution to long-term risk prediction for advanced and interval cancers is unknown.”
Study Details
The study included 2,412 patients with invasive breast cancer who had two-dimensional full-field digital mammograms performed 2 to 5.5 years before cancer diagnosis and 4,995 controls matched for age, race, and date of mammogram from two U.S. mammography cohorts (San Francisco Mammography Registry and Mayo Clinic screening cohort). Data were collected for BI-RADS density, AI detection system score (Transpara version 1.7.0; score = 1–10, representing increasing probability of malignancy at time of breast screening), and volumetric density measurements. Odds ratios (ORs) in analyses adjusted for age and body mass index and C-statistics (AUC) were used to estimate the association of the AI score with invasive cancer and its contribution to models including breast density measures.
Key Findings
On mammograms conducted between 2 and 5.5 years prior to cancer diagnoses, each one-unit increase in AI score was associated with an increased risk of invasive breast cancer (OR = 1.20, 95% confidence interval [CI] = 1.17–1.22; AUC = 0.63, 95% CI = 0.62–0.64), interval cancer (OR = 1.20, 95% CI = 1.13–1.27; AUC = 0.63, 95% CI = 0.59–0.67) advanced cancer (OR = 1.23, 95% CI = 1.16–1.31; AUC = 0.64, 95% CI = 0.60– 0.68) and cancer in dense breasts (OR = 1.18, 95% CI = 1.15–1.22; AUC = 0.66, 95% CI = 0.64–0.67).
AI score improved prediction of all cancer types in models with density measures (likelihood ratio test P values < .001). Discrimination was improved with the addition of the AI score for advanced cancer, with AUC for dense volume increasing from 0.624 to 0.679 (difference = 0.065, P = .01), but change in AUC did not reach statistical significance for interval cancer.
The investigators concluded, “AI imaging algorithms coupled with breast density independently contribute to long-term risk prediction of invasive breast cancers, in particular, advanced cancer.”
Celine M. Vachon, PhD, of the Division of Epidemiology, Department of Quantitative Health Sciences, Mayo Clinic, Rochester, is the corresponding author for the Journal of Clinical Oncology article.
Disclosure: The study was funded by grants from the National Institutes of Health. For full disclosures of the study authors, visit ascopubs.org.
