Construction Of A Machine Learning Prediction Model For Upper Urinary Tract Infected Stones Based On Non-enhanced CT Images

Objectives: This study aims to establish a prediction model of upper urinary tract infectious stones by mining the radiomics features and deep transfer learning features of CT images of upper urinary tract stones using machine learning methods,and to establish a alignment diagram combined with clinical features to provide a decision basis for the auxiliary diagnosis and personalized treatment of upper urinary tract infectious stones in vivo.Methods: Preoperative CT plain images and other clinical data of 780 patients with upper urinary tract stones,including 165 infected stones and 615 non-infected stones,whose stone composition was determined by infrared spectroscopy at the Second Affiliated Hospital of Kunming Medical University from January 2016 to December 2021 were retrospectively analyzed.The CT plain scan images of stones were manually segmented and extracted radiomics(Rad)features,while deep transfer learning(DTL)features of CT images of stones were extracted using the pre-trained resnet34 algorithm,and t-test,Spearman rank correlation test and Least absolute shrinkage and selection operator(Lasso)regression were used for the fusion features of clinical,Rad,DTL features and images Rad-DTL features for feature selection,and then train machine learning classification models such as support vector machine(SVM),K-nearest neighbor(KNN),random decision forest(RF)and XGBoost to build classification models separately and compare their performance by determining the area under the curve(AUC).The accuracy of the three models,Rad,DTL and Rad-DTL,as well as the AUC,were compared,and the best performing one was selected,and alignment diagram were created by combining the classification models built with clinical features.The area under the ROC curve(Receiver Operating Characteristic Curve)was used for model evaluation,the calibration curve was plotted to evaluate the compliance of the model,and the decision curve analysis(DCA)was plotted to evaluate the clinical usefulness of the model.Results: The 1218 and 512 image features were extracted by radiomics and deep transfer learning,respectively.10 Rad features,21 DTL features and 41 "Rad-DL" features were retained after feature screening,and 7 clinical risk factors with P < 0.05 were retained after clinical feature screening.The accuracy and AUC of the Rad feature model were 82.1% and 0.763,respectively,and the accuracy and AUC of the DTL feature model were 80.7% and 0.806,respectively,and the accuracy and AUC of the Rad-DTL feature model were 87.2% and 0.630,respectively.The accuracy,AUC of the classification model built with clinical features was 80.8% and0.630,respectively.by comparison,the optimal model for image features was Rad-DTL,and the AUC of the alignment diagram built with clinical features was0.917(95% CI: 0.850-0.985).Both the calibration curve and the decision curve of the alignment diagram achieve better performance.Conclusions: The upper urinary tract infective stone prediction model established by fusing Rad features,DTL features,and clinical features can successfully predict infective and non-infective stones in vivo preoperatively,and the alignment diagram model we developed can be used as a non-invasive tool to identify upper urinary tract infective stones in vivo to optimize disease management of urolithiasis and improve patient prognosis.