Automatic Assignment of Radiology Examination Protocols Using Pre-trained Language Models with Knowledge Distillation

Selecting radiology examination protocol is a repetitive, and time-consuming process. In this paper, we present a deep learning approach to automatically assign protocols to computed tomography examinations, by pre-training a domain-specific BERT model (BERT_rad). To handle the high data imbalance across exam protocols, we used a knowledge distillation approach that up-sampled the minority classes through data augmentation. We compared classification performance of the described approach with n-gram models using Support Vector Machine (SVM), Gradient Boosting Machine (GBM), and Random Forest (RF) classifiers, as well as the BERT_base model. SVM, GBM and RF achieved macro-averaged F1 scores of 0.45, 0.45, and 0.6 while BERT_base and BERT_rad achieved 0.61 and 0.63. Knowledge distillation boosted performance on the minority classes and achieved an F1 score of 0.66.