The Application Study Of Dynamic ¹¹C-Acetate PET/CT Imaging In The Diagnosis Of Liver Tumor

In the clinical application,The diagnosis of the liver tumor plays an important role in the prevention and treatment of the liver tumor.The ¹¹C-acetate applied in the liver with dynamic PET/CT is an effective way to diagnose the liver tumor.In this study,a compartment model which described the metabolic process of the ¹¹C-acetate in liver was established base on the dynamic PET/CT image.The kinetic parameters in the compartment model was used as the evaluation criteria for the the diagnosis of the liver tumor.In the preprocessing of the dynamic PET/CT images,an optimized time sampling plan was proposed.The time sampling plan divided the dynamic PET/CT images into 25 frames,which included 12 frames of 10 s image,6 frames of 20 s image,4 frames of 30 s image,2 frames of 60 s image and 1 frames of 120 s image.With this plan,the time activity curve of ¹¹C-acetate became more smoothness.In the process of fitting the input function,we used a three exponential function which could simulate the variation features of the input function appropriately.Furthermore,a method which used the slope threshold of the time activity curve was used to cut off the meaningless dynamic PET/CT images.As a result,the total weighted residual of the fitting function decreased 27%.The liver has two blood supply,which included the hepatic artery and the portal vein.In this study,we built an one compartment model to simulate the two blood supply,and at the same time,we also built the three compartant model for the metabolic process of the ¹¹C-acetate in liver.A parameter estimation algorithm called the pixel based graphed nonlinear least square algorithm（PGNLS）was proposed to estimate the kinetic parameters in the dynamical model above.The PGNLS algotithm could estimated the parameters pixel by pixel,this feature allowed us to acquire the distribution map of these kinetic parameters.the simulation results showed that the PGNLS algorithm ensured the accuracy of the estimated kinetic parameters and improved the estimation reliability when compared with the conventional GNLS algorithm.In real patient data test,the proposed method substantially improved the contract between tumor and normal tissue by an average of 54.31 percent,which made the small size tumor more obviously and provided more details information of the large size tumor.To identify the appropriate parameters which could differentiate the normal tissue,the benign lesion and the malignancy,the Wilks’ Lambda test was used to test each kinetic parameter.（?） and（?） was chosen finaly.Then the linear discriminant functions were constructed by the Fisher criteria.The lesion was classified into the group whose discriminant function has the highest score,and the accuracy of three group was 100%,71.4%,75%,respectively.From the patient data,we found that va still couldn’t differentiate the benign lesion and the malignancy at some situation.Therefore,we introduced two new parameters: （?）and（?）.These two parameters were nonlinear and had a wider value range which made them more sensitive in classification.The classification result showed that （?）improved the accuracy of the malignancy,which is 87.5%,and（?）improved the accuracy of the benign lesion,which is 85.7%.