Automated ICD-10 code assignment of nonstandard diagnoses via a two-stage framework.

An electronic medical record (EMR) is a rich source of clinical information for medical studies. Each physician usually has his or her own way to describe a patient's diagnosis. This results in many different ways to describe the same disease, which produces a large number of informal nonstandard diagnoses in EMRs. The Tenth Revision of International Classification of Diseases (ICD-10) is a medical classification list of codes for diagnoses. Automated ICD-10 code assignment of the nonstandard diagnosis is an important way to improve the quality of the medical study. However, manual coding is expensive, time-consuming and inefficient. Moreover, terminology in the standard diagnostic library comprises approximately 23,000 subcategory (6-digit) codes. Classifying the entire set of subcategory codes is extremely challenging. ICD-10 codes in the standard diagnostic library are organized hierarchically, and each category code (3-digit) relates to several or dozens of subcategory (6-digit) codes. Based on the hierarchical structure of the ICD-10 code, we propose a two-stage ICD-10 code assignment framework, which examines the entire category codes (approximately 1900) and searches the subcategory codes under the specific category code. Furthermore, since medical coding datasets are plagued with a training data sparsity issue, we introduce more supervised information to overcome this issue. Compared with the method that searches within approximately 23,000 subcategory codes, our approach requires examination of a considerably reduced number of codes. Extensive experiments show that our framework can improve the performance of the automated code assignment.

Chinese clinical named entity recognition with radical-level feature and self-attention mechanism.

Named entity recognition is a fundamental and crucial task in medical natural language processing problems. In medical fields, Chinese clinical named entity recognition identifies boundaries and types of medical entities from unstructured text such as electronic medical records. Recently, a composition model of bidirectional Long Short-term Memory Networks (BiLSTMs) and conditional random field (BiLSTM-CRF) based character-level semantics has achieved great success in Chinese clinical named entity recognition tasks. But this method can only capture contextual semantics between characters in sentences. However, Chinese characters are hieroglyphics, and deeper semantic information is hidden inside, the BiLSTM-CRF model failed to get this information. In addition, some of the entities in the sentence are dependent, but the Long Short-term Memory (LSTM) does not capture long-term dependencies perfectly between characters. So we propose a BiLSTM-CRF model based on the radical-level feature and self-attention mechanism to solve these problems. We use the convolutional neural network (CNN) to extract radical-level features, aims to capture the intrinsic and internal relevances of characters. In addition, we use self-attention mechanism to capture the dependency between characters regardless of their distance. Experiments show that our model achieves F1-score 93.00% and 86.34% on CCKS-2017 and TP_CNER dataset respectively.