This multicenter study aims to evaluate whether active breast ultrasound screening can improve the early diagnosis rate of breast cancer in women aged 65 to 80 years, compared with routine elderly health management. The study uses a mixed cluster-individual randomized design based on community implementation capacity. A subgroup of pilot communities will simultaneously collect ultrasound AI data for research performance analysis only, without affecting clinical diagnosis.

This is a multicenter, open-label, mixed cluster-individual randomized controlled trial conducted in Jiangsu Province, China. A total of 82,440 elderly women aged 65-80 years will be enrolled and assigned to either the ultrasound active screening group or the routine health management group. Communities with limited screening capacity will be randomized as clusters, while communities with individual screening conditions will randomize participants individually.

The primary endpoint is the absolute difference in the diagnosis rate of early-stage breast cancer between the two groups within 2 years of follow-up. In pilot communities of the screening group, an ultrasound AI system will be deployed synchronously to verify its diagnostic performance, homogenization effect on primary care physicians, and health economic value. All clinical diagnosis and endpoint adjudication are based on manual ultrasound reports; AI results are only used for retrospective scientific research and do not interfere with clinical decision-making.

Inclusion Criteria:

• Female participants aged 65-80 years.
• Local residents living in participating communities for ≥6 months.
• Signed informed consent and willingness to participate.

Exclusion Criteria:

• History of breast cancer or previous breast cancer treatment.
• Breast cancer screening within the past 2 years.
• Severe comorbid conditions that may affect study participation or follow-up.
• Acute breast disease or conditions requiring immediate treatment.
• Unable to cooperate with screening procedures due to physical or cognitive limitations.

Individual participant data will not be shared because there is currently no approved data sharing policy for this multicenter study. De-identified participant data will be managed in accordance with institutional regulations and applicable ethical and legal requirements.