Physiological Model Constrained PET Reconstruction

Positron Emission Tomography, known as PET, is the most advanced nuclear medical imaging modality which can provide functional images in vivo. By detecting the radioactive signals inside human body, PET can record the position of annihilation coincidence and save the data as a Sinogram map in the computer. Mathematical methods are performed on the Sinogram to recover the activity map, which indicates the metabolic changes of the organs.With the continuous development of PET technology. the key element of concern, PET reconstruction, has gained increasing attention. Reconstruction methods including analytical method, iterative method, as well as the state space method are proposed in successive. In order to meet the ever-increasing demands of PET imaging, this paper is devoted to study the existing problems in the reconstruction field. The following issues are covered to further explore the reconstruction problem under physiological model constraints.Firstly, a robust state space model is built to replace the current widely-used model which simplifies the system error and noise property. By taking these issues into account, the new model is much more suitable for realistic PET application. With the constraints of the physiological model, a robust adaptive Kalman filter is performed to jointly estimate the radioactivity and the noise statistical parameters. Experiments show that the new method achieves better results under the new model compared with traditional MLEM and FBP methods.Secondly, this paper initiates the research topic of limited view PET reconstruction. Traditional PET has been hindered to use in a large scale by its high cost and large amount of data operation needs. While the Dual-Head Tomography with Coincidence system is applicable, there are several challenges with regard to hardware precision and image resolution. The limited view reconstruction is of great interest in both terms of detector cost and data processing. Furthermore, it lays the foundation for the development of dual-head Tomography with Coincidence system. The tracer kinetic constrained framework is constructed to solve the limit view PET reconstruction problem, where a large amount of data is missing. Computer simulation and real data experiments have proved the feasibility of this method.