| ObjectiveEvidence-based Medicine has advanced the practice of clinical trials in Traditional Chinese Medicine,and numerous clinical trial literatures have been published by TCM researchers.However,most TCM clinical trial literatures present knowledge in textual,tabular,and graphical forms,and knowledge resources are scattered and isolated among different literatures.Traditional ways of acquiring knowledge require searching,downloading,and reading from electronic literature databases,which is laborious and time-consuming.Moreover,when synthesizing medical evidence,data must be extracted manually from the papers,which is inefficient.In this situation,it is hard for computers to comprehend and utilize unstructured evidence-based data,and clinical researchers cannot obtain clinical evidence quickly and efficiently.To address the current issues of the lack of efficient channels for obtaining clinical evidence and the inability to integrate and share clinical trial knowledge,the study proposes a framework for constructing a knowledge graph of TCM clinical trials by integrating multiple literature sources,which is based on the structural and linguistic characteristics of TCM clinical trial literature.Using empirical research on TCM clinical trial literature data for diabetic kidney disease,the study organizes the discrete knowledge resources in the literature into an interconnected knowledge system.Building on the knowledge extraction task,the study follows strict evaluation principles,combines methodological and reporting quality,and achieves automatic bias-of-risk assessment and literature quality scoring,providing high-quality evidence for clinical researchers.Method(1)Based on literature research,ontology and non-ontology resources such as TCM Language System(TCMLS),Scientific Evidence and Source Information Ontology(SEPIO),CONSORT 2010 Statement and Cochrane ROB(Bias-Of-Risk)Tool were used to identify core concepts such as disease,evidence,symptoms,grouping,interventions,outcome indicators and literature quality information.Semantic associations between interdisciplinary concepts in TCM and evidence-based medicine were established and a TCM clinical trial knowledge ontology was constructed as a model layer for knowledge graph.(2)Based on resources such as the Terminology of Traditional Chinese Clinical Diagnosis and Treatment,Classification and Codes of Traditional Chinese Medicine Diseases and Syndromes,and Evidence-Based Clinical Practice Guidelines for TCM Treatment of Diabetes Mellitus,a terminology dictionary was constructed to improve the accuracy of term extraction,and a list of synonymous terms was constructed to standardize term expressions and match different referent terms of entities.To address the problems of nested entities and overlapping relationships in TCM clinical trial literature,a joint entity relationship extraction model based on a multi-headed selection framework was trained to conduct an empirical study using diabetic kidney disease TCM clinical trial literature as the data source,extracting entities and relationships from it,and storing and visualizing the data using Neo4j after knowledge fusion.(3)Based on the knowledge extraction task,entities related to the quality of the literature can be obtained,and the study of literature quality evaluation is carried out in terms of both automatic bias-of-risk assessment and literature quality scoring.The Word2Vec model is used to represent the bias-of-risk judgment standard statements and the bias-of-risk description statements extracted from the literature,and the biasof-risk judgment is achieved by calculating the semantic similarity between them.In addition,combining with methodological quality and report quality,the data elements that can characterize the quality of the literature are selected to construct scoring rules considering the factors of data representability,citation and readability,and the literature quality score is calculated to provide a ranking basis for the intelligent Q&A of the knowledge graph.Results(1)A knowledge ontology with well-expressed knowledge was constructed.The TCM clinical trial knowledge ontology clarified the hierarchical knowledge concepts in related fields,identified categories such as basic literature information,clinical data information,statistical analysis information,trial design information and methodological quality information.Semantic associations between TCM concepts such as disease,syndrome and symptoms and evidence-based medicine concepts such as intervention measures and outcome indicators were established.A total of 68 classes,8 object properties and 38 data properties were constructed.(2)A TCM clinical trial knowledge graph that supports evidence-based comprehensive evaluation of diabetic nephropathy was constructed.In the data layer construction,5856 formula terms,1168 proprietary Chinese medicine terms,4197 traditional Chinese medicine terms,2641 syndrome terms and 5476 Western medicine terms were organized to improve the accuracy of term extraction.The positive and negative noun table constructed included a total of 16168 terms to standardize term expression and reduce data redundancy.A total of 1893 literature documents on TCM clinical trials for diabetic nephropathy were extracted,and a total of 14462 entities and 194799 relationships were obtained through knowledge fusion.(3)The developed TCM clinical trial knowledge service platform has embedded an automatic literature quality evaluation function that can be used for evidence synthesis and screening of high-quality evidence.Based on Word2Vec bias-of-risk assessment and rule-based literature quality scoring,the platform has implemented automatic bias-of-risk judgment through similarity matching.It can generate a biasof-risk assessment diagram required for system evaluation in the platform.By scoring the quality of literature according to rules,it can provide users with high-quality results in intelligent question-answering services.Conclusion(1)The construction of the TCM clinical trial knowledge ontology has achieved structured expression and interdisciplinary knowledge organization of TCM clinical trial literature knowledge.Semantic associations between TCM concepts and evidence-based medicine concepts have been established,laying the foundation for subsequent data annotation and knowledge extraction tasks.(2)The knowledge graph of TCM clinical trials for diabetic kidney disease constructed based on the ontology framework converts the literature from a humanreadable format to a machine-readable form,achieving the ordered and associated expression of TCM clinical trial knowledge and facilitating the integration and sharing of clinical trial knowledge.(3)The automatic bias-of-risk assessment and literature quality scoring based on knowledge extraction are explorations of evidence-based practice with computable evidence,which can improve the efficiency of evidence synthesis and meet the needs of clinical researchers to obtain high-quality evidence quickly.Highlight(1)Conceptual association of interdisciplinary knowledge organization.To overcome the disciplinary knowledge islands that hinder knowledge integration and sharing among different disciplines with distinct research paradigms and knowledge systems,the study exploratively establishes a conceptual association between traditional Chinese medicine(TCM)and evidence-based medicine(EBM),two fields that are richly represented in the literature of TCM clinical trials.It enables the construction of a knowledge system within the literature of TCM clinical trials.(2)Automatic literature quality evaluation.From the perspectives of bias-of-risk assessment and literature quality scoring,the study implements an automatic evaluation of TCM clinical trial literature on the TCM clinical trial knowledge service platform built by the project.It can not only automatically generate bias-of-risk graphs required for systematic reviews,but also score literature according to selfmade rules,providing a basis for ranking automatic question answering results. |
PDF Full Text Request