Bundled AI supported automated breast ultrasound versus handheld ultrasound in organized screening in China.
Authors
Affiliations (6)
Affiliations (6)
- Department of Cancer Epidemiology, National Cancer Center/National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing 100021, China.
- Department of Surgery, Maternity and Child Healthcare Hospital of Luohu District Shenzhen City, Shenzhen 518019, China.
- Breast and Thyroid Surgery Department, Maternity and Child Healthcare Hospital of Longgang District, Shenzhen City (Longgang Maternity and Child Institute of Shantou University Medical College), Shenzhen 518060, China.
- Breast and Thyroid Surgery Department, Maternity and Child Healthcare Hospital of Futian District, Shenzhen 518000, China.
- School of Biomedical Engineering, Shenzhen University Medical School, Shenzhen University, Shenzhen 518060, China.
- Department of Ultrasound, Xijing Hospital, Fourth Military Medical University, Xi'an 710032, China.
Abstract
Evidence for AI-supported ultrasound screening programs remains limited. We compared a bundled AI-supported automated breast ultrasound (ABUS) workflow with handheld ultrasound (HHUS) in a government-organized screening program. Adjusted risk ratios accounted for baseline imbalance, and standardized cancer detection rate (CDR) accounted for differential biopsy completion. ABUS reduced recall (12.1% vs. 29.9%; adjusted RR 0.45 [95% CI 0.42-0.48]) and biopsy recall (0.99% vs. 2.19%; adjusted RR 0.50 [0.39-0.64]), while increasing biopsy positive predictive value (PPV; 62.1% vs. 20.0%; adjusted RR 2.46 [1.34-4.52]). Observed CDR was higher but not statistically significant (3.08 vs. 1.22 per 1000; adjusted rate ratios 1.90 [0.94-3.86]). Early-stage proportion was higher with ABUS but not statistically significant (84.8% vs. 50.0%; adjusted RR 1.64 [0.71-3.80]). The bundled ABUS pathway was associated with substantially lower recall and higher biopsy PPV. Findings for cancer detection and stage distribution were exploratory because of differential biopsy completion and the small number of cancers.