Search

Background We aimed to build a common terminology in the domain of cervical cancer, named Cervical Cancer Common Terminology (CCCT), that will facilitate clinical data exchange, ensure quality of data and support large scale data analysis. Methods The standard concepts and relations of CCCT were collected from ICD-10-CM Chinese Version, ICD-9-PC Chinese Version, officially issued commonly used Chinese clinical terms, Chinese guidelines for diagnosis and treatment of cervical cancer and Chinese medical book Lin Qiaozhi Gynecologic Oncology. 2062 cervical cancer electronic medical records (EMRs) from 16 hospitals, belong to different regions and hospital tiers, were collected for terminology enrichment and building common terms and relations. Concepts hierarchies, terms and relationships were built using Protégé. The performance of natural language processing results was evaluated by average precision, recall, and F1-score. The usability of CCCT were evaluated by terminology coverage. Results A total of 880 standard concepts, 1182 common terms, 16 relations and 6 attributes were defined in CCCT, which organized in 6 levels and 11 classes. Initial evaluation of the natural language processing results demonstrated average precision, recall, and F1-score percentages of 96%, 72.6%, and 88.5%. The average terminology coverage for three classes of terms, clinical manifestation, treatment, and pathology, were 87.22%, 92.63%, and 89.85%, respectively. Flexible Chinese expressions exist between regions, traditions, cultures, and language habits within the country, linguistic variations in different settings and diverse translation of introduced western language terms are the main reasons of uncovered terms. Conclusions Our study demonstrated the initial results of CCCT construction. This study is an ongoing work, with the update of medical knowledge, more standard clinical concepts will be added in, and with more EMRs to be collected and analyzed, the term coverage will be continuing improved. In the future, CCCT will effectively support clinical data analysis in large scale.

Background: Early prediction of the clinical outcome of patients with sepsis is of great significance and can guide treatment and reduce the mortality of patients. However, it is clinically difficult for clinicians.Methods: A total of 2,224 patients with sepsis were involved over a 3-year period (2016–2018) in the intensive care unit (ICU) of Peking Union Medical College Hospital. With all the key medical data from the first 6 h in the ICU, three machine learning models, logistic regression, random forest, and XGBoost, were used to predict mortality, severity (sepsis/septic shock), and length of ICU stay (LOS) (>6 days, ≤ 6 days). Missing data imputation and oversampling were completed on the dataset before introduction into the models.Results: Compared to the mortality and LOS predictions, the severity prediction achieved the best classification results, based on the area under the operating receiver characteristics (AUC), with the random forest classifier (sensitivity = 0.65, specificity = 0.73, F1 score = 0.72, AUC = 0.79). The random forest model also showed the best overall performance (mortality prediction: sensitivity = 0.50, specificity = 0.84, F1 score = 0.66, AUC = 0.74; LOS prediction: sensitivity = 0.79, specificity = 0.66, F1 score = 0.69, AUC = 0.76) among the three models. The predictive ability of the SOFA score itself was inferior to that of the above three models.Conclusions: Using the random forest classifier in the first 6 h of ICU admission can provide a comprehensive early warning of sepsis, which will contribute to the formulation and management of clinical decisions and the allocation and management of resources.

We present a post-processing technique for correcting misrecognition results derived from a continuous speech recognizer with more sophisticated linguistic knowledge. Firstly, we propose a novel confidence measure called word viability that is based on word activation force (WAF) model, which has the ability to correct some syntactic and semantic errors by means of taking the information about the neighbors into consideration. Secondly, the error correction algorithm corrects recognition errors by first creating candidate list for errors, and then re-ranking the candidates with a combination of mutual information score, trigram score and word viability score. Mandarin film experiments show that the combination of mutual information, trigram and word viability can effectively reduce the character error rate of speech recognition.