Objective:

To evaluate the intrasystem repeatability of the S-Detect CAD system when identical stored breast ultrasound images were reanalyzed under standardized blinded conditions.

Approach:

• Study Design: A retrospective, registered, blinded repeatability study approved by an ethics committee.
• Participants: 261 female patients with breast nodules were included after screening 269 patients.
• Image Acquisition: Ultrasound images were obtained using a standardized protocol on a Samsung Medison RS80A system.
• S-Detect Analysis: Two independent analyses (S-Detect 1 and S-Detect 2) were performed on the same images with a minimum 4-week interval.
• Reference Standard: Histopathological diagnosis served as the reference standard for lesion classification.
• Statistical Analysis: Concordance and repeatability were quantified using Cohen κ, and diagnostic performance metrics were calculated.

Key Findings:

• 39.2% of nodules were classified as 'possibly benign' and 60.8% as 'possibly malignant' in the first analysis.
• Repeatability between S-Detect 1 and S-Detect 2 was assessed, with discrepancies indicating system-level variability.

Interpretation:

Limitations:

• The study was conducted at a single center, which may limit generalizability.
• The analysis was based on static images, which may not reflect dynamic clinical scenarios.

Conclusion:

The findings provide insights into the repeatability of the S-Detect CAD system in breast ultrasound classification.