Using ANN And Serum Protein Pattern Models In Liver Cancer Diagnosis

Backgroud : For a long time, the liver cancer comes to the second place in the mortality rate in the malignancies in China, just behind the lung cancer. It is not difficult in the diagnosis of the primary liver cancer with the typical signs and symptoms. Hence, liver cancer tends to present at a late clinical stage with poor prognosis. The AFP (alpha fetoprotein) has been a gold standard for diagnosis of the liver cancer in early stage. The AFP is a kind of protein produced by liver cells and yolk sac of the fetus which will disappear one week after the birth. After canceration, the liver cell is capable of producing AFP again. According to the pathological grading of liver cancer, well-differentiated and poor-differentiated liver cancer cells do not produce the AFP. Only the moderate-differentiated liver cancer cells can produce the AFP. So the positive rate of the AFP in diagnosis of liver cancer ranges from 60% to 70%. And there are also high positive rate of AFP among the patients with liver cirrhosis . It is indicated that there is a problem of false positive when the AFP is applied in the diagnosis of primary liver cancer. So it is urgent to find out a simple technique of early diagnosis with high sensitivity, specificity and quickness, particularity at the clinical medical ascend.The component and quantity of proteins in the cells will change prior to the occurrence of the pathological changes of malignant tumor. So in theory, it is possible to screen the index and signs of diseases in early stage by dynamic observation of the protein. The precondition is the specific molecule for certain disease for early diagnosis. The screening method is high-throughput and cannot be performed with the traditional techniques. The research and application of proteomics along with its related high technology make this large scale screening come to truth. It also opens a fresh new stage of the molecular diagnosis technology for the clinical application, especially for the tumor diagnosis.This research adoption the United States ciphergen SELDI- TOF- MS to analysis contrast the serum protein of liver cancer, normal person and cirrhosis, this method sensitive and specific, intend to proceed examination to test this method whether can used to choose the specific serum protein marker of the earlier period of liver cancer.Purpose: Establish a technique to examinate the serum protein mass spectometry, and study the value of serum protein mass spectrometry according to artificial nerve the network model in the diagnosis of the liver cancer.Material and method: 106 serum samples obtained, 52 were from patients with liver cancer, 22 from patients with liver cirrhosis in the Department of Oncology of the Second Affiliated Hospital of Zhejiang University, China, and 32 from healthy volunteers in the same hospital. The diagnosis of the liver cancer was confirmed postoperatively by pathological examination in all 52 patients including 32 males and 20 females with an average age of 60.6 years old (39-81 years old). All samples including the 22 liver cirrhosis and 32 healthy controls were age and gender matched, which were obtained in the early morning under the condition of no eating and drinking and stored at -80℃ in a cryogenic refrigerator after separation. And the serum samples from the patients with liver cancer were collected prior to the start of treatment. This research used SELDI- TOF- MS to analysis contrast the serum protein of liver cancer, normal person and cirrhosis. At thesame time, 106 example specimens is divided into two set in radiam: train set(70) and test set(36), firstly train 70 example specimens proceeds artificial nerve network, get a model, then examinate this model. Secondly,use the same technique and m/zta collect method to estimate 36 unknown serum specimen( include the liver cancer, the cirrhosis and the health person). Use same method to collect the serum protein mass spectometry of 36 examples , get the spectra m/zta equally to the train set,proceeding matrix converting after output the primitive m/zta, then...