A BERT-Based Artificial Intelligence to Analyze Free-Text Clinical Notes for Binary Classification in Papillary Thyroid Carcinoma Recurrence

Patient information in free text form exists in medical information systems. Before the successes of the natural language processing models, it had costed resources to refine unstructured information into neat information formats for training artificial intelligence models. Here, we applied the bidirectional encoder representations from transformer (BERT) classifier to analyze unstructured clinical text information on diagnosis of the recurrent papillary thyroid cancer (PTC). It showed a neat performance of 98.8% in the binary classification of PTC recurrence.