A Radiomics Nomogram Model For Predicting Prognosis Of Pancreatic Cancer After High-Intensity Focused Ultrasound Surgery

Objective:The purpose of this study was to develop a nomogram based on MRI radiomics in the training cohort to predict the prognosis of patients with pancreatic cancer after highintensity focused ultrasound treatment,and to further verify it in the validation cohort.Methods:A retrospective analysis of 52 patients with pancreatic cancer who underwent HIFU in the Affiliated Hospital of Qingdao University from July 2019 to May 2021 was performed.The patients were divided into a training cohort of 32 patients and a validation cohort of 20 patients.Linear correlation analysis was performed using Pearson correlation analysis.The features with correlation coefficient greater than 0.9 are eliminated.Useful radiomics features were then selected using a 10-fold cross-validated least absolute shrinkage selection operator Cox regression method.Radiomics model,clinical model and radiomics nomogram model were constructed by multiple regression analysis.The identification,calibration and clinical application of nomogram were further evaluated.Potential association between Rad-Score and prognosis was assessed by K-M survival analysis.Results:Through lasso regression analysis,11 radiological features(7 from CET1-w arterial phase imaging and 4 from CET1-w portal venous phase imaging)were selected to construct radiomics features.ROC curves were further used to evaluate the performance of the developed radiomics features in two cohorts.The diagnostic efficacy of Rad-Score was worthy of recognition,with AUC values of 0.722(training cohort)and 0.705(validation cohort),respectively.According to the radiomics score,the patients were divided into highrisk and low-risk groups.K-M analysis showed that the difference between the two groups was statistically significant(p < 0.05).The prognosis of patients in the low Rad-Score group was better than that in the high Rad-Score group.In addition,Cox risk regression showed that Rad-Score and tumor size were independent factors for OS.Rad-Score was significantly associated with overall survival(P < 0.001).Compared with clinical models and radiomics models,the radiomics nomogram constructed by combining Rad-Score and clinicopathological factors can better predict the survival of patients.The calibration curve shows that the model predictions of 6-month,9-month and 1-year OS have the best consistency with actual observations.DCA analysis showed that the radiomics nomogram model had a larger area under the curve than the clinical model and the radiomics model,indicating that the radiomics nomogram is a more practical clinical tool for predicting the prognosis of pancreatic cancer patients.Conclusion:Rad-Score is a significant prognostic factor for pancreatic cancer patients.In addition,the combined nomogram based on Rad-Score and clinicopathological factors has great potential in predicting OS of patients,which can help us better predict the survival of patients after HIFU,thus providing help for clinical practice.