An agentic, no-code artificial intelligence workflow for developing a thyroid nodule ultrasound malignancy classifier
Authors' conclusion
An agentic, no-code AI workflow produced a calibrated, externally validated thyroid nodule classifier, supporting accessible, reproducible, and independently testable medical AI development.
Study context
- Population
- Thyroid nodules on US; internal test set from TN5000 dataset; external cohort of nodules from the University of Colorado spanning multiple cancer histologies
- Participants
- 232 external validation nodules
- Comparison
- Agentic no-code artificial intelligence (AI)-developed ResNet-18 classifier vs previously published benchmark classifier
- Outcome
- Diagnostic performance for thyroid nodule malignancy on external validation
Methodological detailsretrospective, computational, external validation
- Retrospective
- Computational
- External validation
Performance
As reported in the abstract
| Test or model | AUROC | Sensitivity | Specificity | PPV | NPV |
|---|---|---|---|---|---|
| Agentic model | 0.90 (95% CI, 0.85-0.93) | 0.92 | 0.68 | 0.52 | 0.96 |
| Benchmark model | 0.83 | — not stated in the abstract | — not stated in the abstract | — not stated in the abstract | — not stated in the abstract |
AUROC
Agentic model0.90 (95% CI, 0.85-0.93)
Benchmark model0.83
Sensitivity
Agentic model0.92
Benchmark model— not stated in the abstract
Specificity
Agentic model0.68
Benchmark model— not stated in the abstract
PPV
Agentic model0.52
Benchmark model— not stated in the abstract
NPV
Agentic model0.96
Benchmark model— not stated in the abstract
— not stated in the abstract
- Agentic model vs Benchmark model: paired DeLong P = .03
Further findings
| Outcome | Agentic model |
|---|---|
| Diagnostic performance on internal test set | |
| AUROC | 0.94 (95% CI, 0.92-0.95) |
| Sensitivity | 0.90 |
| Specificity | 0.81 |
Missed cancers
- Missed cancers clustered among follicular pattern tumors
Original publication
An agentic, no-code artificial intelligence workflow for developing a thyroid nodule ultrasound malignancy classifier. J Clin Endocrinol Metab. Published 29 September 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →