The Value Of CT Imaging In Distinguishing Pulmonary Cystic Echinococcosis From Pulmonary Abscess

Purpose: Objective: To investigate the value of noninvasive CT imaging analysis in differentiating lung cystic echinococcosis(CE)rupture from lung abscess.Methods:From January 2010 to November 2020,40 patients(22 males,18 females)with lung CE rupture confirmed by pathology in our hospital were retrospectively analyzed.Meanwhile,clinical data of 80 patients(44 males,36 females)with lung abscess examined during the same period were selected according to a ratio of 1:2.The CT images were observed and delineated by random double blind method.A total of 131 VOI lesions were delineated,and then a random number of patients was generated using a computer-generated 6:2: 2.Divided into training set(72 persons),verification set(24 persons)and test set(24persons),Lasso algorithm was used to screen out the best features,and four different classifiers(KNN,LR,RF and SVM)were used to perform machine learning on the extracted features.The efficiency of the models constructed by each classifier was evaluated by the four main indexes of area under ROC curve(AUC),sensitivity,specificity and accuracy.Results: The consistency test showed good consistency between the two,ICC > 0.75.After dimension-reduction with Lasso algorithm,the 1409 features extracted from CT images were screened out and 19 optimal features(3 first-order features,5 three-dimensional texture features and 11 high-order features)were selected.The model constructed by the two classifiers KNN and SVM based on CT images had the best performance.The AUC value,sensitivity and specificity of differentiating patients with CE rupture from lung abscess using KNN were 0.832,0.909,and 0.615,respectively.The AUC value and sensitivity of differentiating patients with CE rupture from lung abscess using SVM were 0.832,0.909,respectively.The specificity was 0.692,and the performance of SVM classifier was better through the analysis of four classifiers,including accuracy,recall rate,F1 score and support degree.Conclusion: The use of CT imaging tomics to extract the texture features of valuable lesions can make up for the deficiency of visual observation,and it is of great significance in the identification of ruptured lung CE and lung abscess.