PET System Response Modeling

Position emission tomography(PET)is a powerful molecular imaging modality by measuring the radiotracer distribution in vivo,which is widely used in clinical applications.At the same time,PET imaging continue to be influenced by various physical degradation factors.The use of a statistical iterative reconstruction algorithm with a physically accurate system modeling leads to improved spatial resolution and image quality.However,it is a challenging problem to model PET system fast and accurately.Monte Carlo methods,which can simulate physical factors in PET imaging,is widely used to model PET system response.However,the methods need a significant computational cost as well as a large memory capacity to store files.To address the problem,here we improve the method of using symmetries by simulating an axial wedge region.This approach takes full advantage of intrinsic symmetries in the cylindrical PET system without significantly increasing the computation cost in the process of symmetries.A total of 4224 symmetries are exploited.The Aliyun elastic compute service is employed to calculate the simulation,which enables us to access powerful computing capabilities.And it takes about two weeks to finish system modeling.A large number of system matrix files increase the reconstruction time.In this paper,the traditional OSEM algorithm is improved by rotating images during PET reconstruction to take full advantage of symmetries in PET system.In this way,the size of matrix files is reduced to 13.2GB,about 1/3899 of total matrix files.Multithreading method are used to further accelerate image reconstruction.In order to accomplish a more flexible system modeling method,the feature of PET system response is studied carefully.The gaussion function is used to model system blurring in radial and axial directions.The parameters of gaussion function are derived by processing 151 point source data at different position.Various phantoms are expoited to evaluate the performance of different system response modeling methods.The studies show Gaussion function based system response modeling method can effectively reduce the spatial variance of resolution in field of view,and improve the image contrast,as well as the Monte Carlo based system response modeling.