Prediction Of Acute Myocardial Infarction Disorders Based On Topic Model And Prediction Of Trends

[Purpose] Based on the clinical text data of acute myocardial infarction,this study used data analysis and mining technology to construct acute myocardial infarction disease recognition and disease change treatment mode,and found that the current electronic medical record clinical text data problems and give some suggestions,the resulting model can provide decision support for clinical diagnosis and treatment,promote the exploitation of medical and health big data,and also be clinical text data mining analysis provides new ideas.[Methods] Based on the electronic medical record data of acute myocardial infarction extracted from a regional health information in Yichang,Hubei Province,the data obtained are classified according to different disease nodes.Pre-processing of texts such as the deciphering,sorting and jieba word segmentation are targeted according to whether the overall data structure is standard.And we can discover the problems existing in the current clinical text data during the pre-processing process and make corresponding suggestions.Then use the topic model LDA to generate the theme of the clinical text in the electronic medical record,and use the minimum principle of the confusion function to get the best theme output result.Then the results of each disease node were analyzed and synthesized.According to the time of admission-hospital-discharge,the results were analyzed to form the disease recognition and disease change treatment mode of acute myocardial infarction,and the text similarity was calculated by JS distance to identify and predict the disease.Finally,through the comparison with the standard diagnosis,clinical path,different disease types,word segmentation tools to adjust the results and the generated patterns,verify the accuracy,rationality,usability,for the diagnosis of acute myocardial infarction and development and relevant treatment measures prediction recommendations.[Results] Through continuous experimental improvement,a more effective treatment mode for acute myocardial infarction disease recognition and disease change has beengenerated.Compared with the standard diagnosis of the disease,the hospital-generated disease generated by the study is more universal and reasonable,and can effectively match the actual disease text.Compared with the results of the data output of the word segmentation and the undifferentiated disease node without adding the external dictionary,the results obtained by adding the dictionary and the diseased node are more interpretable and reduce the ambiguity problem.The model was tested using the acute myocardial infarction test set data with an accuracy rate of 89%,with old myocardial infarction and the accuracy of the data test models of myocardial infarction recovery and coronary atherosclerotic heart disease were 71% and 68%,respectively.The results showed that the model can better distinguish acute myocardial infarction and other different diseases.Compared with the clinical path of standard acute myocardial infarction and the disease recognition and change treatment mode,the generated model is reasonable,and the model includes the main body of the complaint and other aspects not mentioned in the clinical path.[Conclusions] The clinical text data mining analysis in the Chinese field is still in its infancy,and it is very urgent to mine high-value data information in clinical texts.This study used the topic model to generate patterns for the identification and prediction of acute myocardial infarction in the validation test.Compared with similar studies in the past,the output results are more accurate and reasonable,but there is room for improvement.At the same time,it also provides new ideas for Chinese clinical text mining.