Deep Learning Based Temporal Information Extraction Framework on Chinese Electronic Health Records

Abstract

Electronic Health Records (EHRs) are generated in the clinical treatment process and contain a large number of medical knowledge, which is closely related to the health status of patients. Thus information extraction on unstructured clinical notes in EHRs is important which could contribute to huge improvement in patient health management. Besides, temporal related information extraction seems to be more essential since clinical notes are designed to capture states of patients over time. Previous studies mainly focused on English corpus. However, there are very limited research work on Chinese EHRs. Due to the challenges brought by the characteristics of Chinese, it is difficult to apply existing techniques for English on Chinese corpus directly. Considering this situation, we proposed a deep learning based temporal information extraction framework in this paper. Our framework contains three components: data preprocessing, temporal entity extraction and temporal relation extraction. For temporal entity extraction, we proposed a recurrent neural network based model, using bidirectional long short-term memory (LSTM) with Conditional Random Fields decoding (LSTM-CRF). For temporal relation extraction, we utilize Convolutional Neural Network (CNN) to classify temporal relations between clinical entities and temporal related expressions. To the best of our knowledge, this is the first framework to apply deep learning to temporal information extraction from clinical notes in Chinese EHRs. We conduct extensive sets of experiments on real-world datasets from hospital. The experimental results show the effectiveness of our framework, indicating its practical application value.