Predicting excellent response to radioiodine in differentiated thyroid cancer using machine learning
1Department of Nuclear Medicine, Recep Tayyip Erdogan University, Faculty of Medicine, Training and Research Hospital, Rize, Turkey.
Summary
Machine learning models can predict excellent response (ER) after radioactive iodine (RAI) treatment for differentiated thyroid carcinoma (DTC) patients. This prediction aids in assessing recurrence risk in DTC without distant metastasis.
Area of Science:
- Oncology
- Medical Imaging
- Machine Learning
Background:
- Differentiated thyroid carcinoma (DTC) patients achieving excellent response (ER) after radioactive iodine (RAI) treatment exhibit a low recurrence rate.
- Predicting ER early is crucial for managing DTC patients without distant metastasis.
Purpose of the Study:
- To develop and evaluate machine learning (ML) models for predicting ER at 6-24 months post-RAI treatment in DTC patients.
- To identify key clinicopathological parameters influencing ER prediction.
Main Methods:
- Utilized clinicopathological data, including thyroidectomy/neck dissection pathology, laboratory results, and imaging findings from 151 DTC patients without distant metastasis.
- Applied various ML models to predict ER status (ER/nonER) based on pre-treatment and post-RAI data.
Main Results:
- 118 patients achieved ER, while 33 had nonER post-RAI.
- Thyroglobulin antibodies (TgAb) were significantly more prevalent in the nonER group before RAI (55% vs. 29%, p=0.007).
- Eight ML models demonstrated high predictive performance (AUC > 0.700), with extreme gradient boosting achieving the highest AUC (0.871) and gradient boosting showing 81% accuracy.
Conclusions:
- ML models show significant potential for accurately predicting ER in DTC patients treated with RAI.
- These predictive models can assist clinicians in risk stratification and treatment planning for DTC patients.
