Predicting Early Liver Metastasis After Pancreatic Cancer Surgery Based On Enhanced CT Radiomic And Clinical Factors

OBJECTIVE: To establish a comprehensive prediction model based on enhanced CT Radiomic combined with clinical factors and to evaluate its value in predicting the risk of early postoperative liver metastasis in pancreatic cancer.METHODS: A retrospective analysis of 112 patients admitted to the First Hospital of the University of Science and Technology of China from January 2018 to May 2021 with postoperative pathologically confirmed pancreatic ductal adenocarcinoma(PDAC)in parallel with preoperative enhanced CT examination and postoperative with a complete follow-up processwas performed,and the patients were randomly divided into a training group(n=79)and a validation group(n=33)in a 7:3 ratio.General clinical data and surgical pathology information of all patients were collected,and independent clinical predictors were screened using univariate and multivariate Cox regression analyses,and a clinical prediction model for early liver metastasis after pancreatic cancer surgery was established.The 3D images of the arterial and venous phases of the enhanced CT lesions of each patient were outlined layer by layer semi-automatically using ITK-SANP(www.itksnap.org)software,and the original imaging histological features of the pre-processed lesions were extracted using A.K Software(A.K Software,version 3.3.0.R;GE Healthcare)software,and the intraclass correlation efficiency(ICC)and Least Absolute Shrinkage and Selection Operator(LASSO)regression analysis was used to screen out stable Radiomic features,calculate the Radiomic score(Radsocre),and establish an Radiomic prediction model for early liver metastasis after pancreatic cancer surgery.Finally,the independent clinical predictors were combined with arterial and venous phase Radiomic features to establish a comprehensive model of early liver metastasis after pancreatic cancer surgery,which was presented visually as a column line graph(Nomogram)and validated internally in the validation group.The efficacy of the models was assessed using consistency indices(C-index),and the clinical utility of the models was assessed using decision curve analysis(DCA).Each patient was scored using the three models separately,divided into high-risk and low-risk groups based on median values,and K-M survival curves were plotted,and the Log-rank method was used to test the differences between the two groups.RESULTS: Serum amylase(HR,1.001;95%CI,1.000-1.002;P=0.009)and CEA(HR,3.522;95% CI,1.763-7.038;P < 0.001)were screened as independent clinical predictors of early liver metastasis after pancreatic cancer surgery by univariate and multivariate Cox regression analysis,and the C-index of the clinical model established by the two clinical factors was 0.75(95% CI,0.68-0.82)and 0.75(95% CI,0.61-0.89)in the training and validation groups,respectively,A total of 1050 original Radiomic features were extracted by A.K software analysis,eight arterial phase and five venous phase Radiomic features were extracted after feature Screening,and the C-index of the constructed Radiomic model was 0.81(95% CI,0.73-0.88)and 0.72(95% CI,0.57-0.87)in the training and validation groups,respectively.Combining the Radiomic features with independent clinical factors to construct a combined model,the final combined model had a C-index of 0.83(95% CI,0.76-0.89)and 0.82(95% CI,0.69-0.94)in the training and validation groups,respectively.the DCA showed a higher net benefit of the joint model than the other two groups in most of the threshold range.CONCLUSION: The combined model established by Radiomic combined with clinical factors can better predict the risk of early postoperative liver metastasis in pancreatic cancer and stratify patients by risk,thus helping clinicians to develop individualized treatment plans and guide patient prognosis.