In breast cancer diagnostics, distinguishing suspicious findings from those requiring further investigation remains a challenge. Dense breast tissue can make abnormalities difficult to identify, while different imaging methods provide complementary information.
Researchers at Sun Yat-sen Memorial Hospital, Sun Yat-sen University, and collaborating institutions developed an AI model combining ultrasound and digital breast tomosynthesis (DBT) to support breast cancer risk assessment.
The model achieved a specificity of 95.5 percent and a sensitivity of 85 percent in an independent pathology-confirmed cohort of 500 breasts. By combining complementary information from both imaging modalities, the model could provide radiologists with additional support when assessing suspicious or inconclusive findings.
The model's improved specificity could help radiologists distinguish findings that warrant further investigation from those that may not require the same level of follow-up.
However, the researchers emphasize that the model is intended to complement radiologists' assessments. Prospective multicenter validation is still required before clinical implementation.