A Preliminary Study On Predicting The Biological Boundary Of Hepatic Alveolar Echinococcosis Lesions Based On MRI Imaging Group

Objective:To explore the application value of the imaging omics model based on T2-weighted magnetic resonance images to predict the biological boundary of hepatic alveolar echinococcosis(HAE)in preoperative surgery.Methods: A retrospective collection of 89 patients with HAE confirmed by pathology after operation in our hospital was conducted.All patients had conventional abdominal MRI plain scan images before operation,and the invasion of the marginal area of HAE lesions was determined according to the postoperative pathological results.Labelers were established in the invasion group(32 cases)and non-invasion group(57 cases).They were randomly assigned to the training set(n= 70)and the test set(n=19)in an 8:2 ratio.Firstly,the Radcloud platform was used to delineate the ROI of HAE lesion layer by layer from T2 WI sequence images and to extract the imaging omics features of the lesion.The initial results showed that there were many imaging omics features,most of which were not related to the pathological results.Then,two methods,variance threshold method and univariate selection method,were used for characteristic selection.The dominant signs associated with HAE microvascular invasion were screened out.The features and pathological information were extracted from the above dimensionality reduction,and then three classifiers,Random Forest(RF),Extreme Gradient Boosting(XGBoost)and Logical Regression(LR),were used to construct the prediction model.And the modeling efficiency of each classifier was evaluated by three main indexes of area under the ROC curve(AUC),sensitivity and specificity and two secondary indexes of accuracy and recall rate.Results: 1409 relevant features were extracted from T2-weighted magnetic resonance images,and 7 optimal image omics features were selected by dimensionality reduction to build the model,among which the XGBoost classifier had the best performance.The AUC value,precision,sensitivity and specificity of the training set corresponding to the HAE edge invasion group were 0.96,0.82,0.92 and 0.89,respectively,and the AUC value,precision,sensitivity and specificity of the test set were 0.89,0.83,0.71 and 0.92,respectively.Conclusion: The radiomics model based on MRI T2 WI can be used as a potential biomarker to predict the biological boundary of HAE and provide a new basis for clinical radical hepatectomy.