Based On Radiomics:Prognsis Of Patients With Acute Paraquat Poisoning

Objective: To establish and evaluate the efficiency of radiomics model in predicting prognosis of patients with acute paraquat poisoning(APP).Methods: The data was gathered from November 2014 to October 2017 that consisted of chest computed tomography(CT)images of 80 patients with APP early-mid term and related clinical data,which were randomly assigned to a training group and a verification group according to 7: 3 ratio.The training group(57 cases)was used to establish the prediction model,and the verification group(23 cases)was used to validate the model.The whole lung was chosen as the region of interest(ROI)when the developing peak of lung lesion,the radiomics features were extracted from ROI..Principal component analysis(PCA)and lasso regression was used to reduce the data dimension,select features and establish radiomics signature(radscore).The multivariate logistic regression analysis was used to establish a radiomics comprehensive prediction model,which included radscore and clinical risk factors that is blood laboratory text indexes,and the model results were represented by nomogram.The performance of the nomogram was assessed with respect to its discrimination,calibration,and clinical usefulness.Result: The radscore,which consisted of 7 selected features,was a statistically significant difference(P <0.001),for both training dataset and validation dataset).The area under the ROC curve of operation(AUC)was0.942(95% CI 0.886-0.997)and 0.865(95% CI 0.658-1)respectively in the training dataset and validation dataset.,and the sensitivity and specificity respectively were 0.864,0.914 and 0.778,0.929,the prediction accuracy rates respectively were 89.5% and 87%.Predictors included in the individualized predictive nomograms include radscore,PQC,CK-MB,and SCr.The AUC of nomogram was 0.973(95% CI 0.936-1)in training dataset,and the sensitivity and specificity were 0.943,0.955 respectively,the prediction accuracy was94.7%.This predictive model showed good discrimination and good calibration.AUC also reached 0.944(95% CI 0.844-1)in the validation dataset,and sensitivity and specificity were 0.889,0.929 respectively,the prediction accuracy was 91.3%.Decision curve analysis demonstrated that the radiomics nomograms were clinically useful.Conclusion: The radscore could effectively distinguish the prognosis of APP patients.hematological laboratory datas increasesd the prognostic value of nomogram.The comprehensive model improved the accuracy of early individualized prediction of APP patients.It was helpful to accurately assess the severity in early stages of poisoning and reliably predict the risk of death.It could be uesed to guide the adjustment of clinical treatment programs,in order to reduce mortality and disability rates,and reduce the medical costs of patients.This study also provides a theoretical basis for application of radiomics in non-neoplastic and diffuse lesions.