Recognition Of The Non-Traumatic Benign And Malignant Vertebral Fractures In Middle-Aged And Elderly People Based On CT Radiomics

PurposeConstruct a CT-based radiomics feature model,combined with clinical features,to jointly construct an image-clinical joint model,which is used to evaluate the diagnostic efficiency of the model for benign and malignant vertebral fractures,and better assist clinicians in formulating next-step treatment strategies.Materials and MethodsPatients who underwent spinal CT examination(cervical spine CT plain scan/thoracic spine CT plain scan/lumbar spine CT plain scan)in the Department of Radiology,China-Japan Union Hospital of Jilin University from October 2018 to October 2022 were retrospectively collected(benign and malignant fractures were each 50 cases),collected relevant clinical characteristics(age,gender,and history of malignant tumors)and imaging images,selected thin-layer(1.00mm)serial images,used semi-automatic delineation method to delineate vertebral bodies and accessories,extracted radiomics features,and imaged Features and clinical features The Mann-Whitney U test and the least absolute shrinkage and selection operator(LASSO)algorithm were used for data dimensionality reduction and feature screening.The image features with the strongest correlation with the clinical features were established using the support vector machine(SVM)modeling method to establish a classification model,and the area under the receiver operating characteristic(ROC)curve(area under the curve,AUC),accuracy,macro average,recall rate and F1 value to evaluate the diagnostic performance of the model.Result86 initial features were extracted from the thin-slice sequence images,and the Mann-Whitney U test and LASSO algorithm were used to screen the image features and clinical features to select 10 image features and 1 clinical feature that were highly correlated with the assessment of benign and malignant vertebral fractures.Features,based on the above-mentioned relevant features,a diagnostic model was constructed in the SVM classifier.ROC analysis showed that the AUC values of the training set and the test set were 0.96,0.87,the accuracy was 0.93,0.83,the macro average was 0.93,0.83,F1 The values are 0.93 and 0.83 respectively,which have good classification and diagnosis performance.ConclusionThis study established and verified a more accurate CT-based radiomics model,combined with clinical features,to predict benign and malignant non-traumatic vertebral fractures.By reducing the variability produced by subjective visual analysis and providing quantitative information to improve the diagnostic accuracy and efficiency of radiologists,it provides a basis for further clinical treatment of patients.