AI decision support for breast ultrasound could curb benign biopsies

Among 271 BI-RADS 4 breast ultrasound lesions, an FDA-cleared AI tool classified 60.6% of benign lesions as BI-RADS 2 or 3, potentially reducing benign biopsies. However, 9.5% of malignant lesions were downgraded, warranting careful integration with radiologist judgment.

• The study retrospectively analyzed 271 biopsied BI-RADS 4 lesions (253 women, mean age 54); 23.2% (63/271) were malignant.
• The AI tool downgraded 60.6% (126/208) of benign lesions to BI-RADS 2 or 3, suggesting potential to avoid many benign biopsies.
• However, the AI classified 9.5% (6/63) of malignant lesions as BI-RADS 2 or 3; no prospective follow-up confirmed safety, underscoring the need for cautious implementation.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.