Dynamic PET Image Reconstruction Based On Regional Spatial-temporal Prior

As an outstanding representative of functional molecular imaging modalities, positron emission tomography (PET) provides functional information of physiological activity by displaying the concentration distribution of radioisotope labeled tracer (chemical compounds or biological molecular) which pre-injected into the human body before imaging process, and has shown great performance in oncology, cardiopathy, neurology and new medicine studies by representing heart and brain metabolism and functions on molecule level.Especially in clinical applications, with the development of radiopharmaceutical chemical and hardware devices, PET has been widely used in clinical diseases diagnosis, and been the important imageology diagnostic tool. By means of the specificity of imaging radiopharmaceuticals, PET shows its irreplaceable advantages in the clinical diagnosis, such as cancer, heart disease, neurological and psychiatric diagnosis, and is more and more concern and attention. Meanwhile, in the eyes of clinical doctors, PET has been the gold standard for further diagnosis and treatment.However, positron emission tomography is an ill-posed inverse problem because of the low counts rate and the low signal to noise ratio of the observed projection data which are contaminated by noise and physical effects. Though needing less computation cost, traditional filter back projection (FBP) method often reconstruct noisy images of low quality. Better expressing system models of physical effects and modeling the statistical Poisson character of the data, the famous maximum likelihood expectation maximization (ML-EM) approach outperforms the FBP method with regard to image quality. However, pure traditional ML-EM approach suffers slow convergence and the reconstructed activity images always start deteriorating to produce "checkerboard effect" as the iteration proceeds.Works on high quality PET reconstruction are mainly concentrated in the following two strategies. One is reconstructing PET images with the MAP (maximum a posteriori) approach by incorporating prior information, while the other is improving the quality of reconstructed PET images by means of various image denoising techniques. MAP reconstruction which is also called Bayesian reconstruction based on MRF (Markov random fields) theory, incorporates MRF prior information of objective tracer concentration distribution data into the ML-EM algorithm through regularization or prior terms and has been proved theoretically correct and practically effective. While works of denoising to improve image quality are usually by use of sinogram recovery and image post-processing techniques, according to the statistical properties of sinogram data and noise characteristics of the reconstructed image. Sinogram recovery by means of mathematically modeling the measured data's statistical properties, in general, Poisson statistical model approximately, use the appropriate denoising method for recovery. While image post-processing is to reduce the artifacts noise mainly.Our work on the high quality reconstruction algorithms of PET image are as following:(1) For Bayesian dynamic Positron emission tomography (PET) reconstruction, a regional spatial-temporal prior (RST) model is proposed in this paper. RST exploits sufficiently both the spatial and temporal information, and could suppress the noise in both two spatial dimensions and one temporal dimension. Computer simulations of brain 18F-FDG dynamic PET was conducted to validate the proposed approach. The comparison with other classical reconstruction methods shows that the proposed approach performs better in dynamic PET reconstruction and results in more accurate estimate of the influx rate of 18F-FDG in the lesion region.(2) A novel approach is proposed to improve PET image quality by a non-local means prior induced by projection recovery. The new method exploits the redundancy of information in the image reconstructed from the restored projection data by optimizing non-local weight. First, the raw projection data is restored, and the PET image is reconstructed with conventional filtered back projection (FBP). A weight matrix is calculated by the non-local means weighting formula, and then the direct reconstructed image is restored with a non-local means prior guided by this weight matrix. Simulated and clinical PET data experiment results demonstrate that the proposed method has an effective quality improvement to the PET image under the condition of meeting the clinical needs.