Clinical Value Of Detection Of Patients With Pancreatic Cancer Serum Tumor Markers

Background:Pancreatic cancer is a lethal disease of digestive system. Unfortunately, most people are discovered so late that there are only 15-20% patients of with pancreatic cancer have the opportunity to be operated. Because pancreas locates deeply in abdomen and biological behavior of pancreatic cancer is especial, in early stage of pancreatic cancer, there is few specific symptoms. It is very difficulty to diagnose pancreatic cancer in its early stage. Serum tumor markers take the hope of the early pancreatic cancer diagnoses because their detection with simple method, low cost, and quickly result responding. The effort to find the new serum tumor markers with high specificity and sensitivity is never stopped. However, none of existing serum tumor markers can satisfy the clinical need solely. Because the limitation of the traditional combinational examination, how to build a new and effective model is a dilemma to the investigator. In recent, there is a spring of the combination between life sciences, bioengineering and computer technology. We can apply the fuzzy cluster analysis method and artificial neural networks (ANN) theory to the early diagnosis of pancreatic cancer with computer technology, and then improve the clinical value of serum tumor markers. Objectives:In this study, we employed the technology of ELISA and immunohisto -chemical method to evaluate the clinical value of human OPN and RCAS1 in early diagnosing of pancreatic cancer. We establish a new-type method in diagnosis pancreatic cancer with combinational examination of serum tumor marker using fuzzy cluster analysis method and artificial neural networks (ANN) theory, to evaluate they diagnostic value and set up application program.Method and Result:PartⅠThe Diagonal Value of RCAS1 and OPN in pancreatic cancerThe concentration of RCAS1 and OPN in serum was examined by ELISA, and then analyzed the results by ROC curve. The samples include 46 pancreatic cancer patients, 18 chronic pancreatitis patients, 20 normal persons. Serum levels of RCAS1 and OPN in pancreatic cancer patients is higher than that in chronic pancreatitis patients and normal control obviously (86.2±6.36 vs. 21.51±4.93, 14.40±2.54; 504.51±56.12 vs. 212.87±28.48, 96.32±26.46), P＜0.05. In pancreatic cancer patients, serum OPN level in the tumor can not be exairesised is higher than that in which can be exairesised, and serum RCAS1 in the group with obstructive jaundice is higher than that without obstructive jaundice. There is statistically significant differences, P＜0.05. Compared OPN and RCAS1 with the CA19-9, CA242, CA50 and CEA, ROC curves of the serum tumor markers were made and the AUC (area under curve) was calculated. The AUC ofRCAS1, OPN, CA19-9, CA242, CA50 and CEA are 0.827, 0.813, 0.805, 0.737, 0.530 and 0.727, respectively. The expression of RCAS1 and OPN in pancreatic cancer tissue, chronic pancreatitis tissue and normal pancreas tissue was determined by immunohistochemistry. The results show that the positive rate of expression of RCAS1 and OPN in pancreatic cancer patients is higher than those in chronic pancreatitis patients and normal person.PartⅡThe Diagnostic Value of Artificial Neural Network in pancreatic cancer213 pancreatic cancer and 86 chronic pancreatitis were collected. About 70% of all cases were selected randomly for training samples, and the other 60 cases for testing sample. Firstly we filtrated the 23 variation of diagnosis factors including serum tumor markers CA19-9 etc by using single factor analysis. The results show that there is different between pancreatic cancer patients and chronic pancreatitis patients in sex, age, CA19-9 etc, P＜0.05. The variation of CA19-9, CA242, CA50 and CEA were deal by fuzzy cluster analysis and fuzzy identification method, and then evaluated the veracity of this method and set up application software. The results were compared with Logistic regression obtained by SPSS. Aiming at the same database, CA19-9, CA242, CA50, CEA, sex and age were observed and saved as variation. The trained ANN model (BP neural networks model) was used to test the other 60 cases, on which were compared with Logistic regression obtained by SPSS. The sensitivity and specificity of the fuzzy cluster analysis model (98.67%; 78.57%)is all better than those of Logistic regression model(93.33%; 14.29%), P＜0.05. The diagnosis exactitude rate and AUC of ANN model (89.6%; 0.932) is higher than those of Logistic regression (84.34%, 0.897), P＜0.05.Conclusions:1. The diagnosis ability of RCAS1 and OPN is better than CA50 and CEA conspicuous. If examined with CA19-9 and CA242, it could be improved the diagonal standard of the diagnosis of pancreatic cancer. And RCAS1 and OPN highly expressed in pancreatic cancer tissue.2. The fuzzy cluster analysis model and ANN model is superior to traditional logistic model in the diagnosis of pancreatic cancer. The application software of ANN model is running stably and interface is friendship.