Radiomics Analysis Of Pancreatic Ductal Adenocarcinoma

Background and purpose:Pancreatic ductal adenocarcinoma(PDAC)is a kind of pancreatic malignant tumor with poor prognosis.The current incidence is increasing year by year,which has a great impact on human health.Due to the lack of typical clinical symptoms in the early stages,most PDACs are diagnosed at advanced stages,and the resection rate is low.In addition,because PDAC has the characteristics of high invasiveness,high metastasis and drug resistance,a small number of patients who receive surgical treatment also have a high recurrence rate and do not respond well to follow-up radiotherapy and chemotherapy.Therefore,in order to further improve the therapeutic effect of PDAC and improve the quality of life of patients,it is necessary to look for biomarkers related to the prognosis of PDAC as future treatment targets.And detect it in a non-invasive way as much as possible.Previous studies have shown that the expression levels of C-MYC and HMGA2 proteins are significantly related to the prognosis of many kinds of malignant tumors.Radiomics is a new discipline with rapid development in recent years,which can extract a large amount of information reflecting the biological characteristics of tumor molecules from traditional images.Therefore,in this study,the postoperative pathological specimens of patients with PDAC were used to evaluate the difference of C-MYC and HMGA2 protein expression between tumor tissue and peritumoral tissue,and the relationship between C-MYC,HMGA2 protein expression and patient prognosis and clinical features.Secondly,we radiomics analyzed the preoperative enhanced CT images of PDAC patients,extracted a large number of radiomics features,and used the method of machine learning to classify the features to predict the expression of tumor biomarkers and the prognosis of patients.In order to achieve non-invasive evaluations of biomarkers and prognosis of patients with PDAC,and bring greater hope for clinical treatment.Materials and Methods:1.Analysis of the relationship between the expression of C-MYC and HMGA2 protein in PDAC tissue and the prognosis of patients.This study was approved by the ethics committee of the first affiliated hospital of the army medical university.From March 2013 to September 2015,102 tumor specimens and 93 corresponding peritumoral tissue specimens diagnosed as PDAC were collected retrospectively,and the clinical data of the corresponding patients were collected at the same time.C-MYC and HMGA2 proteins in tumor and peritumoral samples were stained by immunohistochemistry,which were divided into high expression and low expression according to the proportion of positive cells.The difference of C-MYC and HMGA2 protein expression between tumor tissue and peritumoral tissue,and the relationship between C-MYC and HMGA2 protein expression and clinicopathological data and overall survival time of patients were statistically analyzed.2.Radiomics analysis of pancreatic ductal adenocarcinoma.The preoperative enhanced CT images of patients with PDAC diagnosed by surgery from September 2012 to September 2015 were collected,and the portal phase images were used to delineate the region of interest(ROI),and radiomics features(70 conventional features,256 deep features)were extracted.Machine learning is used to classify and model the radiomics features to predict the expression of C-MYC and HMGA2 proteins and the overall survival time of patients.Result:1.Analysis of the relationship between the expression of C-MYC and HMGA2 protein in PDAC tissue and the prognosis of patients.The positive expression rates of C-MYC and HMGA2 proteins in PDAC tissues were significantly higher than those in peritumoral tissues(all p<0.001),and there was a positive correlation between the two proteins(p=0.030).Multivariate logistic regression analysis showed that TNM stage(OR: 5.097,95% CI: 1.546-16.805,p=0.007)and invasion to surrounding tissues and organs(OR: 5.249,95% CI: 1.734-15.886,p=0.003)were independent predictors of C-MYC protein expression,the receiver operating characteristic(ROC)curve shows that the area under the curve(AUC)is 0.8201(95%CI: 0.7345-0.9056,P<0.001).Lymph node metastasis(OR: 4.147,95% CI: 1.653-10.407,p=0.002)and invasion to surrounding tissues and organs(OR: 3.811,95% CI: 1.556-9.336,p=0.003)were independent predictors of HMGA2 protein expression,the receiver operating characteristic(ROC)curve shows that the area under the curve(AUC)is 0.7638(95%CI: 0.6705-0.8572,P<0.001).Lymph node metastasis(RR: 1.632,95% CI: 1.043-2.555,p=0.032),TNM stage Ⅲ + Ⅳ(RR: 1.905,95% CI: 1.139-3.186,P=0.014),positive expression of C-MYC(RR: 1.797,95%CI: 1.101-2.934,P=0.019)and HMGA2(RR: 2.486,95%CI: 1.598-3.866,P<0.001)protein were independent risk factors affecting the survival time of patients with PDAC.2.Radiomics analysis of pancreatic ductal adenocarcinoma.There are 6 radiomics features that are significantly related to the survival of patients with PDAC(p<0.05 and chi-square>3.8).The C-MYC protein expression prediction model based on two deep features showed that the AUC value,accuracy,sensitivity and specificity were 0.90,95%,92% and 98%,respectively.The HMGA2 protein expression prediction model based on one conventional feature showed that the AUC value,accuracy,sensitivity and specificity were 0.91,88%,89% and 88%,respectively.Conclusions:The expression of C-MYC and HMGA2 protein may be important factors affecting the prognosis of patients with PDAC and may become potential therapeutic targets.Based on the radiomics analysis of preoperative enhanced CT in patients with PDAC,the expression of C-MYC and HMGA2 protein in PDAC and the survival time of patients can be predicted more accurately.This non-invasive evaluation method provides potential clinical application value.