Trial shows higher sensitivity with AI-supported breast cancer screening

Artificial intelligence (AI)–supported mammography screening achieved a non-inferior interval cancer rate compared with standard double reading while improving cancer detection sensitivity, according to a large randomized trial published in The Lancet. The findings add evidence that AI can enhance screening performance without increasing missed cancers between screening rounds, according to the Swedish study authors.¹

Studying AI-supported mammography for breast cancers

The population-based trial (NCT04838756) was conducted in Sweden and randomly assigned 105,934 women undergoing routine breast cancer screening to either AI-supported mammography screening or standard double reading without AI.^1,2 Nineteen participants were excluded, leaving more than 105,900 women in the final analysis. Median age was similar between groups at approximately 54 years.¹

In the intervention arm, AI was used both to triage mammograms to single or double radiologist reading and to provide detection support. The primary outcome was interval cancer rate, defined as breast cancers diagnosed between screening rounds or within 2 years after a negative screen, with a prespecified non-inferiority margin of 20%. Secondary outcomes included interval cancer characteristics, sensitivity, specificity, and subgroup analyses by age, breast density, and cancer type.

Interval cancer rates were 1.55 per 1000 participants (95% CI 1.23–1.92) in the AI-supported group and 1.76 per 1000 participants (1.42–2.15) in the standard double-reading group. The proportion ratio of 0.88 met the criterion for non-inferiority (95% CI 0.65–1.18; P = 0.41).

Although the trial was not powered to formally compare interval cancer characteristics, descriptive analyses showed fewer unfavorable cancers in the AI group. Women in the AI-supported arm had fewer invasive interval cancers (75 vs 89), fewer larger tumors classified as T2 or higher (38 vs 48), and fewer non-luminal A cancers (43 vs 59) compared with the control group.¹

AI mammography sensitivity and specificity vs standard double reading

Screening sensitivity was significantly higher with AI support at 80.5% compared with 73.8% with standard double reading (95% CI 76.4–84.2 vs 68.9–78.3 [P = 0.031]). This improvement was consistent across age groups and breast density categories and was observed for invasive cancers, though not for in-situ cancers.¹

Specificity was identical in both groups at 98.5%, with no statistically significant difference (P = 0.88).

“AI-supported mammography screening showed consistently favourable outcomes compared with standard double reading, with a non-inferior interval cancer rate, fewer interval cancers with unfavourable characteristics, higher sensitivity, and the same specificity, while also reducing screen reading workload,” the study authors concluded. “These findings imply that AI-supported mammography screening can efficiently improve screening performance compared with standard double reading and may be considered for implementation in clinical practice.”

References:

1. Gommers J, Hernström V, Josefsson V, 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. The Lancet. Volume 407. Issue 10527. 505 - 514.
2. Mammography Screening With Artificial Intelligence (MASAI) (MASAI). Clinicaltrials.gov. Updated April 4, 2025. Accessed February 2, 2026. https://clinicaltrials.gov/study/NCT04838756