Diagnostic accuracy of an artificial intelligence-based breast ultrasound tool in pregnant and lactating patients
An AI breast ultrasound tool performs like radiologists in pregnant and lactating women, potentially reducing unnecessary imaging.
In 504 pregnant and lactating patients, an AI-based breast ultrasound tool demonstrated sensitivity equivalent to radiologists (80%) and, after excluding lesions outside its intended scope, had comparable biopsy recommendation rates while shifting more lesions to BI-RADS 2 from BI-RADS 3. This suggests AI may reduce unnecessary follow-up imaging in this population while maintaining diagnostic sensitivity.
What the study was
- Study design
- Retrospective diagnostic study (2 academic centers)
- Population
- Pregnant or lactating patients with breast ultrasound findings at MGH and BWH, 2018–2021; n=504 women (639 findings, 5 malignant, 634 benign)
- Sample size
- 504
- Category
- Diagnostics
- Maturity
- Validated
- Journal
- European Radiology
Why it surfaced
Clinically specific AI validation in underrepresented pregnant/lactating population; two academic medical centers; small number of malignancies (5) limits statistical power; practical BI-RADS reassignment implications.
A plain-language summary of published research — not medical advice. Talk to a clinician about your own care.