Analysis Of Radiological Report For Clinical Decision Support

With the development of the integration of the Internet and medical information system technologies,the characteristics of big data in electronic medical record texts have become increasingly significant.The electronic medical record is generated by the doctor's clinical radiation diagnosis and treatment process for patients,and contains a large number of patients' medical activity records.Based on the electronic medical records mining and analysis,it can achieve more accurate risk prediction and patient stratification,further support medical diagnosis and decision support system,and provide auxiliary information for personalized and accurate medical diagnosis.However,the electronic medical record contains a large number of clinical texts,such as radiology reports,pathology reports,etc.,which limits the in-depth mining and application of electronic medical records.This article according to the radiology reports in the clinical record texts,the Conditional Random Field(CRF),Naive Bayes Classification(NBC),Convolutional Neural Network(CNN)and other algorithms are used to analyze information extraction and text classification.Firstly,perform data preprocessing,format conversion,word segmentation,and part-of-speech tagging on radiological reports;According to the experiment needs of this article,with reference to the I2B2(Informatics for Integrating Biology & the Bedside),defining five medical entities:body,check,sign,treatment,and disease,we study the definition and classification methods of entities and entity relationships in radiology reports,and establish a small radiological report annotation corpus.Moreover,we design and implementation of named entity recognition and relationship extraction from radiology reports based on CRF.Secondly,based on the word bag model,feature extraction and representation of radiology reports,and based on this,based on Naive Bayes radiology reports text classification,and analyze the influence of named entities on text feature optimization,Improve radiology reports text classification effect;secondly,based on words The vector represents the words in the report of the reflexology,based on the convolutional neural network(CNN)to extract the features of the radiology text and to achieve the classification of radiological reports,In order to analyze the judgment basis and reference results of the patient's health status.