Radiomics Analyses For Differentiating Pneumonia And Acute Paraquat Lung Injury

Objective:To develop and validate a radiomics nomogram incorporating radiomics signature and laboratory markers,for differentiating bacterial pneumonia and acute paraquat lung injury.Methods: The institutional human research committee approved this retrospective study;informed consent was waived.Patients with pneumonia and acute paraquat who underwent CT examinations between December 2014 and October 2017 were retrospectively identified.Radiomic features were extracted from computed tomography(CT)images of acute paraquat lung injury and pneumonia patients.The study included 180 patients(primary cohort:n=126;validation cohorts:n=54).Lasso regression model was used for dimension reduction,feature selection and radiomics signature building.A prediction model for differentiating bacterial pneumonia and acute paraquat lung injury was built by using backward logistic regression and was presented on a nomogram.Then the prediction model was evaluated with calibration curve and decision curve in primary cohort and validation cohorts.Results: The radiomics signature,which consisted of 34 selected features,was significantly associated with paraquat poisoning and pneumonia(P<.001 for both primary and validation cohorts).The modles showed AUC0.870(95%CI 0.757-0.894)? sensitivity 0.857,specifcity 0.804,positive predictive value83.3% for identifcation,while the validation cohorts showed similar results(0.865(95% CI 0.686-0.907),0.833,0.792,and 81.5%,respectively).The individualized nomogram included radiomics signature,body temperature,nausea and vomiting,and AST.Addition of Clinical risk factors to the nomogram increased the predictive value of the model.The model showd a good discrimination in primary and validation cohorts.Accuracy in training data sets reaches 97.6%.The modles showedAUC 0.897(95%CI 0.821-0.979)?sensitivity 0.900,specifcity 0.958,positive predictive value 92.6% in the validation cohort.The calibration curves and decision curve show that the radiomics nomogram is clinically useful.Conclusion: This study presents a radiomics nomogram that incorporates the radiomics signature and Clinical risk factors,for patients with a fuzzy history of exposure to paraquat poisoning,radiomics can improve radiologists the ability traditional to identify paraquat poisoning lung injury and pneumonia,providing a good opportunity for precision medicine and individualized treatment.