Moderate evidence
More
AI breast ultrasound tool made 40 false alarms vs 14 for standard scans, study of 130 women finds
Study title: Diagnostic performance of S-detect artificial intelligence compared with conventional breast ultrasound for lesion characterization: a prospective lesion-based study
- Evidence type
- Controlled or validated comparison
- Sample
- n = 130
- Sources
- 1 sourcePeer-reviewed
- Published
- 1 min read
- Truvace record
- TRV-2026-2952cite or check it
Study title: Diagnostic performance of S-detect artificial intelligence compared with conventional breast ultrasound for lesion characterization: a prospective lesion-based study
A prospective study of 130 women with 130 target breast lesions compared the S-Detect AI tool with conventional breast ultrasound. Histopathology found 109 benign lesions and 21 malignant ones. Both methods had 90.5% sensitivity, but S-Detect's specificity was 63.3% against 87.2% for conventional ultrasound.
Read the study abstract
Abstract Purpose The aim of this prospective study was to compare the diagnostic performance of conventional breast ultrasound (US) assessment with S-Detect artificial intelligence (AI) for differentiating benign from malignant breast lesions. The study was not designed to assess reduction in biopsy rates. Materials and Methods A total of 130 women with 130 target breast lesions were prospectively evaluated from January 2024 to December 2025. One target lesion was analyzed per patient. Each lesion underwent conv…

The problem
In a prospective study of 130 women with breast lesions, the S-Detect AI tool had 63.3% specificity and 67.7% accuracy versus 87.2% and 87.7% for conventional ultrasound, with 40 false positives against 14.
In brief
S-Detect produced 40 false-positive results compared with 14 for conventional ultrasound, and its accuracy was 67.7% versus 87.7%. A paired comparison showed it generated significantly more positive classifications (McNemar P < 0.001).
The authors say the findings do not show a reduction in unnecessary biopsies, although the study was not designed to measure biopsy rates. They suggest S-Detect could serve as an adjunct to conventional assessment, with its clinical role needing larger prospective reader studies.
Main points
- The study enrolled 130 women between January 2024 and December 2025, analysing one target lesion per patient.
- Histopathology was the reference standard: 109 lesions (83.9%) were benign and 21 (16.1%) malignant.
- Conventional ultrasound with BI-RADS assessment had 90.5% sensitivity, 87.2% specificity, 57.6% PPV and 97.9% NPV.
- S-Detect matched the 90.5% sensitivity but had 32.2% PPV and 97.2% NPV.
- S-Detect gave 40 false positives versus 14 for conventional ultrasound.
The full story
Each lesion received conventional B-mode ultrasound with a final BI-RADS assessment and an S-Detect analysis, and histopathology served as the reference. Sensitivity, specificity, PPV, NPV and accuracy were calculated with 95% confidence intervals.
S-Detect's positive predictive value was 32.2% versus 57.6% for conventional ultrasound, and its accuracy was 67.7% versus 87.7%. Negative predictive value was similar at 97.2% and 97.9%.
The authors conclude S-Detect may be used as an adjunct to conventional assessment, but its clinical role needs evaluation in larger prospective reader studies using predefined management rules.
Sources
1 on file- Egyptian Journal of Radiology and Nuclear Medicinedoi:10.1186/s43055-026-01910-0
Talk it over
S-Detect matched conventional ultrasound at 90.5% sensitivity but produced 40 false positives against 14. How would you weigh catching the same cancers against the extra false alarms for women?
AI breast ultrasound tool made 40 false alarms vs 14 for standard scans, study of 130 women findshttps://truvace.com/item/diagnostic-performance-of-s-detect-artificial-intelligence-compared-with-convent
S-Detect matched conventional ultrasound at 90.5% sensitivity but produced 40 false positives against 14. How would you weigh catching the same cancers against the extra false alarms for women?
Answer in the threadOnly 21 of the 130 lesions were malignant. How does such a small number of cancers limit what the study can say about sensitivity?
Answer in the threadThe authors suggest S-Detect could be used as an adjunct to conventional assessment. If you were a radiologist in this study, how would you use its result alongside your own BI-RADS assessment?
Answer in the thread
If you have more time
The authors call for larger prospective reader studies with predefined management rules. What would such a study need to measure to show whether biopsies go up or down?
Answer in the threadWhich women might be most affected by a false positive from an AI tool, and what would you want to know about how they were told?
Answer in the thread
ace
What readers say