- Design
- Diagnostic accuracy study with model development and community validation
- Population
- 117 community screening participants (153 plaques) in China
- Primary outcome
- Plaque detection and stenosis grading against portable ultrasound
- Effect
- Plaque detection 94.8% vs 87.6%; weighted kappa 0.84 vs 0.49
Researchers built an AI model to sharpen images from handheld ultrasound devices, validated it in 127 hospital patients, then used it in community screening of 117 people in China, comparing against portable ultrasound as the reference.
The enhanced images identified 94.8% of plaques against 87.6% with standard handheld images. Agreement on stenosis grade rose from moderate to strong (weighted kappa 0.49 to 0.84), and sensitivity for plaques with ultrasound features of vulnerability rose from 47.4% to 63.2%, with specificity staying high.
This is a technical accuracy study, not evidence that finding plaques helps people. Major guidelines, including the USPSTF, advise against screening asymptomatic adults for carotid stenosis, because the harms of follow-on procedures can outweigh the benefits. The useful idea for primary care is narrower: where handheld ultrasound is already used, image enhancement may make it more reliable.
- Do not screen asymptomatic adults for carotid stenosis on the strength of this study.
- Manage cardiovascular risk on risk scores, blood pressure, lipids and smoking, not on plaque hunting.
- Refer patients with transient ischaemic attack or minor stroke for urgent carotid imaging.
- Where handheld ultrasound is already in use, AI enhancement may improve image reliability.
Why it matters
Cheap, sharper handheld scans make carotid screening tempting in primary care, but the case for screening was never the image quality.
Don't overread it
Better plaque detection in 117 people does not show that screening prevents strokes.
The statistics, in plain English
Weighted kappa measures agreement beyond chance: 0.49 is moderate and 0.84 is strong. Sensitivity of 63.2% for vulnerable plaques still means more than a third were missed.
Read the rest in the app
You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.
Scan to keep reading on your phone. No account needed to start.
Tomorrow morning, before your first patient
One edition a day for family medicine, written by the desk, every claim tied to its paper. Six minutes.
Get the app — free