New approach of prediction of recurrence in thyroid cancer patients using machine learning

Although papillary thyroid cancers are known to have a relatively low risk of recurrence, several factors are associated with a higher risk of recurrence, such as extrathyroidal extension, nodal metastasis, and BRAF gene mutation. However, predicting disease recurrence and prognosis in patients undergoing thyroidectomy is clinically difficult. To detect new algorithms that predict recurrence, inductive logic programming was used in this study. A total of 785 thyroid cancer patients who underwent bilateral total thyroidectomy and were treated with radioiodine were selected for our study. Of those, 624 (79.5%) cases were used to create algorithms that would detect recurrence. Furthermore, 161 (20.5%) cases were analyzed to validate the created rules. DELMIA Process Rules Discovery was used to conduct the analysis. Of the 624 cases, 43 (6.9%) cases experienced recurrence. Three rules that could predict recurrence were identified, with postoperative thyroglobulin level being the most powerful variable that correlated with recurrence. The rules identified in our study, when applied to the 161 cases for validation, were able to predict 71.4% (10 of 14) of the recurrences. Our study highlights that inductive logic programming could have a useful application in predicting recurrence among thyroid patients.

Automated summary

This study used inductive logic programming to predict recurrence among thyroid cancer patients, identifying three rules that could predict recurrence, with postoperative thyroglobulin level being the most powerful variable. These rules demonstrated high accuracy in predicting recurrence during validation.