Study On TCM Prescription Recommendation Based On Clinical Data

Clinical decision support systems is the core content of the research on medical artificial intelligence.Traditional Chinese medicine(TCM)is a traditional medical system with unique advantages and characteristics in China.Individualized diagnosis and treatment is the main clinical mode based on the information of four diagnostic methods(mainly including symptoms and signs).Therefore,one of the core tasks of artificial intelligence research in TCM is to realize the decision support technology and method for clinical prescriptions based on symptoms and signs as input and prescription information as output.Considering the complexity and individuality of TCM clinical prescriptions,the research on decision support technology for clinical prescriptions has been an unsolved research problem.With the accumulation of a large amount of clinical data with electronic medical record information as carrier,it is possible to recommend individual prescriptions based on large-scale clinical data combined with current machine learning and information recommendation methods.Based on the large-scale clinical medical record data of TCM combined with information recommendation and representation learning,this paper carries out a study on recommendation methods and systems for TCM clinical prescriptions.The main work includes the following aspects.(1)Using the data of 11,379 TCM clinical medical records after desensitization,data preprocessing was performed based on regular expression.A high-quality clinical prescription recommendation data set was formed by batch structured analysis of chief complaints and symptoms in current medical history.The data set mainly contained 11,379 symptoms,11,247 prescriptions and 463 TCMs.It provides an important data base for the research on information recommendation methods for TCM clinical prescriptions.(2)For the need of symptoms and information of prescription in prescription recommendation,a heterogeneous network of symptoms and drugs was constructed.On this basis,feature learning was performed using the network representation learning,and an algorithm for prescription recommendation was developed based on the analysis of categorized formulas and similarity.The experimental results demonstrated that the community-based prescription recommendation method had good performance in prescription recommendation and achieved individualized recommendation in a certain degree.(3)For the application of clinical data preprocessing and TCM prescription recommendation,this paper designed and implemented a prototype system for TCM clinical prescription recommendation.This system,based on Flask framework and Neo4j graph database,realized clinical data query,structured processing of symptoms,and prescription recommendation based on Web.