Risk Stratifying Indeterminate Thyroid Nodules With Machine Learning.

Area of Science:

• Endocrinology
• Oncology
• Medical Informatics

Background:

• Indeterminate thyroid nodules (ITNs) represent up to 30% of cases after fine needle aspiration biopsy.
• Definitive diagnosis of ITNs typically requires invasive surgical pathology.
• Molecular testing offers pre-operative cancer risk stratification but is costly and invasive.

Purpose of the Study:

• To develop and evaluate a machine learning (ML) algorithm for predicting malignancy in ITNs.
• To utilize data from less invasive diagnostic tests for risk stratification.

Main Methods:

• Retrospective study including 355 ITNs from academic and community centers.
• Tested linear, non-linear, and non-linear-ensemble ML classifiers.
• Evaluated classifier performance using 10-fold cross-validation and AUROC.

Main Results:

• A Random Forest classifier achieved 79.1% accuracy, 75.5% sensitivity, and 82.4% specificity.
• The model demonstrated an AUROC of 0.859.
• 171 out of 355 nodules (48.2%) were malignant.

Conclusions:

• ML accurately risk-stratifies ITNs using existing, non-invasive, and inexpensive data.
• ML models offer a cost-effective alternative to molecular testing for ITNs.
• Future research will prospectively evaluate ML combined with expert judgment.