Based On Radiomics:Differentiation Between Mycoplasma Pneumonia And Bacteria Pneumonia

Objective:To explore the value of radiomics in the differential diagnosis of mycoplasma pneumonia and bacterial pneumonia.Materials and Methods:Clinical and imaging data of 100 cases of mycoplasma pneumonia and bacterial pneumonia admitted to The Third Affiliated Hospital of Southern Medical University from January 2018 to December 2019 were analyzed retrospectively.These cases included 60 patients with mycoplasma pneumonia and 40 patients with bacterial pneumonia.Patients were divided into two groups,including training group(70 cases)and test group(30 cases).The patients were randomly divided into training group(70 cases)and test group(30 cases)according to the ratio of 7:3.The region of interest(ROI)of CT image lesions of all patients was outlined and loaded into Artificial Kit analysis software,several radiomics features were extracted.The Least Absolute Shrinkage and Selection Operator algorithm(LASSO)was used for dimensionality reduction,and some radiomics features with the most diagnostic value were finally selected.The estimate values,corresponding coefficients and intercept terms of the radiomics features of 70 patients in the training group were linearly integrated,and then the corresponding Radscore were obtained.In order to establish an radiomics-clinical model nomogram,we used multivariate logistic regression to integrate the Radscore and clinical risk factors.The Receiver Operating Characteristic(ROC)curve was used to evaluate the sensitivity,specificity,and accuracy of the radiomics model and the radiomics-clinical model in the training group and validated by the test group.Finally,each patient’s type of pneumonia was diagnosed based on the radiomics model and radiomics-clinical model,and the diagnostic efficacy of two models were evaluated respetively.Result:There are 396 image texture features extracted from CT images.The LASSO algorithm was used to reduce the dimension,and 10 features with high value were finally screened out.Among them,there are 1 first-order statistic,6 gray co-occurrence matrices and 3 gray running length matrices.Multivariate logistic regression was used to construct the radiomics model and the radiomics-clinical model which integrated Radscores and clinical risks.In the radiomics model,the sensitivity,specificity and accuracy of the training group were 0.762,0.821 and 78.6%respectively,and the Area Under Curve(AUC)was 0.877;the sensitivity,specificity and accuracy of the test group were 0.667,0.750 and 70%,and the AUC was 0.810.In the radiomics-clinical model,the sensitivity,specificity and accuracy of the training group were 0.976,0.714 and 87.1%respectively,and the AUC was 0.905;the sensitivity,specificity and accuracy of the test group were 0.889,0.667 and 80.0%respectively,and the AUC was 0.847.Conclusion:The study which based on CT images has validated that the feasibility of radiomics as a tool for differential diagnosis of mycoplasma pneumonia and bacterial pneumonia.The radiomics model has a good value for differential diagnosis,however,the radiomics-clinical model which integrated clinical risks has higher diffrential diagnostic efficiency.