Predicting Central Lymph Node Metastasis in Papillary Thyroid Carcinoma

A retrospective study of 710 patients (971 lesions) with clinically node-negative T1–T2 papillary thyroid carcinoma, integrating pathological, ultrasound, thyroid function, and systemic inflammatory indicators. The dataset was used to develop an interpretable gradient boosting decision tree model for predicting central lymph node metastasis, achieving an AUC of 0.830 in the test set. The model was authored by Yalin Zhu and last updated on 2026-04-27.

Use Cases

• Developing preoperative prediction models for central lymph node metastasis based on multimodal clinical features.
• Validating the importance of predictors like tumor size, age, and systemic immune-inflammation index (SII) in thyroid cancer.
• Applying SHAP interpretability frameworks to understand feature contributions in a clinical machine learning model.
• Conducting external validation of predictive models on independent temporal cohorts.

Strengths

• Model performance is quantified with an AUC of 0.830 (95% CI: 0.773–0.887) in the test set.
• The study integrated multiple data modalities, including pathological, ultrasound, thyroid function, and systemic inflammatory indicators.
• External validation was performed on an independent temporal cohort of 50 patients, achieving an AUC of 0.800.
• The model identified five independent predictors and one adjunct variable, providing clinical transparency.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset is very small at 11.2 KB, indicating limited scope.

Provenance

Source

figshare

Collection Method

Retrospective analysis of patient records.

Freshness

Last updated 2026-04-27 04:12:03; freshness should be verified.