Combining Image Similarity and Predictive Artificial Intelligence Models to Decrease Subjectivity in Thyroid Nodule Diagnosis and Improve Malignancy Prediction
Authors' conclusion
The AI application showed good performance for sensitivity and NPV between the two datasets and demonstrated potential for 61.5% reduction in the need for fine needle aspiration and strong correlation to ACR TI-RADS. However, PPV was variable between the datasets possibly from variability in image selection and prevalence of malignancy. If implemented widely and consistently among various clinical settings, this could lead to decreased patient burden associated with an invasive procedure and possibly to decreased health care spending.
Study context
- Population
- Thyroid nodules with definitive diagnosis by cytology or surgical pathologynodules
- Participants
- 192 vs 118
- Comparison
- Artificial intelligence (AI) application vs cytology or surgical pathology
- Stanford datasetn = 192
- Private practicen = 118
- Outcome
- AI application prediction of malignancy
Methodological detailsretrospective, external validation
- Retrospective
- External validation
Performance
As reported in the abstract
| Test or model | Sensitivity | Specificity |
|---|---|---|
| Stanford dataset | 1.0 | 0.55 |
Stanford dataset
Sensitivity1.0
Specificity0.55
- Stanford dataset: PPV
- 0.18
- Stanford dataset: NPV
- 1.0
- Stanford dataset: AUC-ROC
- 0.78
- Private dataset: sensitivity
- 0.91
- Private dataset: specificity
- 0.95
- Private dataset: PPV
- 0.8
- Private dataset: NPV
- 0.98
- Private dataset: AUC-ROC
- 0.93
- Private dataset: accuracy
- 0.94
Further findings
Correlation of AI with ACR TI-RADS
- Stanford dataset: ACR TI-RADS based clinical recommendation, polychoric correlation
- 0.67
- Private dataset: ACR TI-RADS based score, polychoric correlation
- 0.94
Original publication
Combining Image Similarity and Predictive Artificial Intelligence Models to Decrease Subjectivity in Thyroid Nodule Diagnosis and Improve Malignancy Prediction. Endocr Pract. Published 8 August 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →