Breast Neoplasms
Conditions
Keywords
Artificial Intelligence, Breast Cancer Screening, Opportunistic Screening, Non-contrast Chest CT, Breast Ultrasound, Health Examination
Brief summary
This study evaluates the clinical utility of a locked chest CT artificial intelligence model for opportunistic breast cancer screening among women undergoing health examinations. The study includes a retrospective validation phase and a prospective single-arm implementation phase. AI analyzes existing non-contrast chest CT images without additional CT examinations. Clinical physicians make further evaluation decisions based on AI outputs, imaging findings, ultrasound results and clinical information.
Detailed description
Retrospective phase: Historical health examination data will be used for offline validation of the locked CT-AI model. Prospective phase: Eligible women undergoing routine health examination will be consecutively enrolled. Participants receive routine chest CT and breast ultrasound. After routine reports are completed and locked, AI analysis is performed. Cases exceeding predefined thresholds are reviewed by clinicians who determine recall decisions. The study evaluates incremental detection value, recall workflow, safety and feasibility.
Interventions
DEVICECT-AI-assisted breast cancer screening
A fixed artificial intelligence model is applied to existing non-contrast chest CT images obtained during routine health examinations to identify and localize suspicious breast lesions and generate a breast cancer risk score and risk category. No additional CT examination is performed for study purposes. Participants meeting predefined AI review criteria are evaluated by trained physicians, who review the original CT images together with the AI output and make the final decision regarding whether additional breast evaluation is recommended. The AI system does not independently diagnose breast cancer or automatically recall participants. Subsequent imaging, biopsy, or treatment is determined according to routine clinical practice and participant preference.
Sponsors
The First Affiliated Hospital with Nanjing Medical University
Study design
Allocation
NA
Intervention model
SINGLE_GROUP
Primary purpose
SCREENING
Masking
NONE
Eligibility
Sex/Gender
FEMALE
Age
18 Years to 80 Years
Healthy volunteers
Yes
Inclusion criteria
* Female participants aged 18 to 80 years. * Undergoing routine health examination. * For the prospective phase, no clear ongoing breast-related symptoms at baseline. * Availability of complete non-contrast chest CT images. In the prospective phase, chest CT must have been scheduled as part of the routine health examination or for another established clinical purpose and must not be performed solely for this study. * Availability of a contemporaneous routine breast ultrasound report and relevant clinical information. * Chest CT image quality adequate for AI analysis. * Availability of an appropriate follow-up pathway through hospital records, pathology systems, cancer registry data, or approved follow-up methods. * For the prospective phase, study information has been provided through an ethics-approved process and the participant has not actively opted out.
Exclusion criteria
* Previous diagnosis of breast cancer, prior treatment for breast malignancy, or breast malignancy already confirmed before baseline. * Pregnancy or breastfeeding. * Incomplete breast coverage on chest CT, severe image artifacts, missing images, or other conditions that prevent valid AI analysis. * Critical baseline or outcome data are substantially incomplete and cannot reasonably be recovered. * No effective follow-up pathway can be established. * For the prospective phase, the participant actively opts out before AI analysis or explicitly declines use of imaging and clinical data for this study.
Design outcomes
Primary
| Measure | Time frame |
|---|---|
| Incremental Breast Cancer Detection Rate of CT-AI | Within 3 months after the index health examination |
| Sensitivity of CT-AI for Breast Cancer Detection in the Retrospective Cohort | Up to 12 months of retrospective outcome ascertainment |
| Overall Breast Cancer Detection Rate of Combined CT-AI and Breast Ultrasound Screening | Within 3 months after the index health examination |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Specificity of CT-AI | Up to 24 months after the index health examination | Specificity of the fixed CT-AI model for breast cancer detection, using final breast cancer status based on pathology and/or clinical follow-up as the reference standard. Unit of measure: percentage (%). |
| Positive Predictive Value of CT-AI-Assisted Recall | Within 3 months after the index health examination | Proportion of participants recalled following CT-AI-assisted physician review who are subsequently confirmed to have breast cancer. Unit of measure: percentage (%). |
| Proportion of Early-Stage Breast Cancers Detected | Within 3 months after the index health examination | — |
| Physician Recall Rate | Within 3 months after the index health examination | Proportion of participants for whom the reviewing physician recommends additional breast evaluation after CT-AI-assisted review. Unit of measure: percentage (%). |
| Breast Biopsy Rate | Within 3 months after the index health examination | Proportion of participants who undergo breast biopsy following the index screening episode. Unit of measure: percentage (%). |
| 24-Month Interval Breast Cancer Rate | Up to 24 months after the index health examination | Number of breast cancers diagnosed during follow-up after the index screening episode among participants without breast cancer detected at the initial screening assessment, reported per 1,000 participants. |
Contacts
CONTACTGe Ma, doctor
PRINCIPAL_INVESTIGATORQiang Ding
The First Affiliated Hospital with Nanjing Medical University
Outcome results
None listed