Machine learning algorithms are comparable to conventional regression models in predicting distant metastasis of follicular thyroid carcinoma
Authors' conclusion
The XGBoost approach was comparable to the conventional LR method for predicting the risk of distant metastasis for FTC.
Study context
- Population
- Patients with newly diagnosed follicular thyroid carcinoma (FTC) in the Surveillance, Epidemiology, and End Results (SEER) database, 2004–2015; 214 (3.68%) with distant metastasis
- Participants
- 5,809 patients
- Exposure
- XGBoost machine learning (ML) model vs conventional binary logistic regression (LR) model for predicting distant metastasis
- Outcome
- AUROC for predicting distant metastasis
Methodological detailsSEER database 2004–2015, prediction model development, 10-fold cross-validation
- SEER database 2004–2015
- Prediction model development
- 10-fold cross-validation
Main finding
As reported in the abstract
AUROC for predicting distant metastasis, after 10-fold cross-validation
| Group | Value | Area under the receiver-operating characteristic (AUROC) |
|---|---|---|
| XGBoost | 0.855 (95% CI 0.803–0.906) | 0.855 |
| LR | 0.845 (95% CI 0.818–0.873) | 0.845 |
Value
XGBoost0.855 (95% CI 0.803–0.906)
LR0.845 (95% CI 0.818–0.873)
Area under the receiver-operating characteristic (AUROC)
XGBoost0.855
LR0.845
Observational data: associations, not effects.
Further findings
| Outcome | XGBoost |
|---|---|
| AUROC of seven ML algorithms in the test set | |
| AUROC | 0.836 |
- AUROC of seven ML algorithms in the test set — AUROCs of the seven ML algorithms
- 0.746-0.836
AUROC of seven ML algorithms in the test set
- XGBoost had the best prediction performance
Original publication
Machine learning algorithms are comparable to conventional regression models in predicting distant metastasis of follicular thyroid carcinoma. Clin Endocrinol (Oxf). Published 25 February 2022. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →