WHILST it has previously been evidenced that AI can improve mammography screening by reducing reading workload and increasing detection, its effect on interval cancers has yet to be investigated. Interval cancers are defined as primary breast cancers diagnosed between two screening rounds or within 2 years after the last scheduled screening that were not previously detected.
Interval Cancers and AI
In their randomised, controlled, non-inferiority, single-blinded, population-based screening accuracy trial, Gommers et al. sought to fill this research gap, by comparing the interval cancer rate in AI-supported mammography screening versus standard double reading without AI.
Participants were allocated to AI-supported mammography screening and standard double reading without the use of AI in a 1:1 ratio. AI was utilised to triage examinations to single or double reading by radiologists. Whilst the primary outcome in this analysis was interval cancer rate, with a 20% non-inferiority margin, secondary outcomes reported in this study were interval cancer characteristics, sensitivity, specificity, and sensitivity by age, breast density, and cancer type.
Gommer’s Findings
Over a total number of 105,934 participants, interval cancer rates were 1.55 (95% CI: 1·23–1·92) and 1.76 (95% CI: 1·42–2·15) per 1000 participants in the intervention and control group respectively, a non-inferior proportion ratio of 0.88 (95% CI: 0·65–1·18; p=0·41). Following analysis, the AI-supported mammography screening group exhibited fewer invasive interval cancers (75 versus 89), T2+ (38 versus 48), and non-luminal A (43 versus 59) than the standard double reading group. Sensitivity was also recorded as higher in the AI-supported group (80.5% [95% CI: 76.4–84.2]) than the control group (73.8% [95% CI: 68·9–78·3]; p=0·031).
The Future of Breast Screening
Overall, Gommer’s study evinces how AI-supported mammography screening shows consistently favourable outcomes when contrasted with standard double reading. This analysis demonstrates how AI-supported mammography screening can improve screening performance and efficiency. This development signposts the possibility of progression regarding breast cancer diagnosis, as, in the future, AI-supported mammography may be considered for clinical implementation following further research.
Reference
Gommers J et al. Interval cancer, sensitivity, and specificity comparing AI-supported mammography screening with standard double reading without AI in the MASAI study: a randomised, controlled, non-inferiority, single-blinded, population-based, screening-accuracy trial. Lancet. 2026;407(10527):505-14.
Featured image: Valerii Apetroaiei on Adobe Stock
