AI-Assisted Sonography for the Diagnosis of Thyroid Nodules.
Javeria Khan
✦ AI-curated · Sources linked
In 30 seconds
This systematic review assessed the diagnostic performance and clinical utility of AI-assisted sonography for thyroid nodules. It included 30 studies published from January 2018 to June 2025 that compared AI models with histopathology or fine-needle aspiration reference standards. Across studies, area under the curve ranged from 0.78 to 0.97, sensitivity from 75.6% to 94.0%, and specificity from 70.4% to 90.7%. Reported clinical benefits included 10% to 45% fewer fine-needle aspirations and 2.5 to 3.1 minutes saved per case in some studies.
Key findings
- Across included studies, area under the curve ranged from 0.78 to 0.97, sensitivity from 75.6% to 94.0%, and specificity from 70.4% to 90.7%.
- Studies reported a 10% to 45% reduction in fine-needle aspirations performed.
- Some studies reported 2.5 to 3.1 minutes of interpretation time saved per case.
- Risk-of-bias assessment classified 53.3% of studies as low risk, 33.3% as moderate risk, and 13.3% as high risk.
Why it matters
Thyroid sonogram interpretation can vary between observers, and the review suggests AI may support more consistent assessment while reducing biopsy use and interpretation time. The reported performance and workflow benefits are promising, but the range across studies and the presence of moderate- and high-risk studies mean results should be interpreted cautiously.
What to keep in mind
The review reports that risk of bias was often related to retrospective design or small sample size. It calls for prospective multicenter trials and standardized reporting.