An Automated Approach for Clinical Quantitative Information Extraction from Chinese Electronic Medical Records

Clinical quantitative information recognition and entity-quantity association from Chinese electronic medical records

“…For the recognition of entity, numeric and unit, as shown in Table 6, the results of our model obtained a precision of 93.81%, a recall of 94.74% and an F1-measure of 94.27%. Compared with the result using rule-based method proposed previously [2] on the same dataset, our model had an improvement of 1.32% in overall F1-measure. In addition, compared to the base Bi-LSTM-CRF model, our proposed model outperformed it by 0.54% improvement in terms of F1-measure.…”

“…To further understand the performance of our model, we compared it with other widely-used baseline methods. For the task of quantitative information recognition, the 9 The root mean squared error and kappa statistic of each classifier performance of our method exceeded base Bi-LSTM-CRF (F1-measure of 93.73%), CRF (F1-measure of 94.08%) and the rule-based method reported in the previous work (F1-measure of 92.95%) [2]. Since the dataset used in this paper contained 1359 EMRs only, we believe that our model could achieve better performance if a larger dataset could be supplied.…”

“…Clinical quantitative information (CQI) is a kind of quantitative information specifically for medical domain. It consists of clinical entities and correlated quantifiable information [2], covering disease diagnoses, clinical detection indicators, and drug doses. The prevalence of digital medical has raised great mass fervor across the medical information science field [3], and considerable researches are engaged in clinical information structuring, such as the clinical named entity recognition [4] and temporal expression extraction [5].…”