A Deep Learning-Based Artificial Intelligence Model Assisting Thyroid Nodule Diagnosis and Management: Pilot Results for Evaluating Thyroid Malignancy in Pediatric Cohorts
Authors' conclusion
AI-Thyroid demonstrated high performance in diagnosing pediatric thyroid cancer.
Study context
- Population
- Two pediatric cohorts with thyroid nodule ultrasonography (US) and histological confirmation at two institutions; mean age 15.5 ± 2.4 years; 103 girls156 nodules
- Participants
- 128 participants
- Comparison
- AI-Thyroid deep learning (DL) model vs radiologist interpretation using Thyroid Imaging Reporting and Data System (TIRADS)
- Outcome
- Diagnostic performance for malignancy (47 of 156 nodules malignant, 30.1%)
Methodological detailsretrospective, two institutions
- Retrospective
- Two institutions
Performance
As reported in the abstract
| Test or model | AUC | Sensitivity | Specificity |
|---|---|---|---|
| AI-Thyroid | 0.913-0.929 | 78.7-89.4% | 79.8-91.7% |
AI-Thyroid
AUC0.913-0.929
Sensitivity78.7-89.4%
Specificity79.8-91.7%
- AI-Thyroid vs ACR-TIRADS: P = 0.016 — highest AUC with AI-Thyroid
- AI-Thyroid vs K-TIRADS: P < 0.001 — highest AUC with AI-Thyroid
Further findings
AUC across image planes and between cohorts
- AUC did not significantly differ across image planes
- all p > 0.05
- AUC did not significantly differ between the two pediatric cohorts
- P = 0.804
Performance by age group (<14 vs ≥14 years)
- No significant differences in sensitivity and specificity between age groups
- all p > 0.05
- AUC higher for patients aged <14 years vs ≥14 years
- all p < 0.01
Original publication
A Deep Learning-Based Artificial Intelligence Model Assisting Thyroid Nodule Diagnosis and Management: Pilot Results for Evaluating Thyroid Malignancy in Pediatric Cohorts. Thyroid. Published 2 June 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →