Performance Evaluation Of The Dual-head PET System And The Calculation Of System Response Matrix Based On Analytical Derivations

Recently,there has been a trend in developing the dual-head positron emission tomography（PET）systems at the lower cost and complexity,at the same time,the dual-head PET systems have open and adustable configuration.The open structure has the potential to promote the emergence and development of many new PET application.In addition,the adjustable geometry makes it great potential for the detection of breast,as these systems can simply and flexibilly adjust the separation between the two detector heads to match the thickness of the breast.Nevetheless,the projection data of the dual-head PET system is incomplete including the missing view and data truncation,which causes the distortions in the reconstructed image.The iterative algorithm has the lesser demand for the completeness of forward data,and it can improve the imaging performce by a accurate system response matrix（SRM）.The dissertation involves the construction of an accurate SRM and the development of a rotatory PET system.Furthermore,we studies the imaging method of dual-head PET system and the performance evaluation of a rotatory prototype system in this dissertation.The main contributions of this dissertation can be summarized as follows:Firstly,we simulated the characteristics of a rotatory dual-head PET system in this dissertation.The influences of different rotation increments were compared and analyzed,which provided the theoretical information for the construction of a prototype system.A reconstruction flow chart was proposed based on a pre-calculated system response matrix（SRM）based on the Monte Carlo（MC）simulation.The SRM made the relationships between the voxels and lines of response（LORs）fixed,therefore we added the interpolation method into the flow chart.Five metrics,including spatial resolution,normalized mean squared error（NMSE）,peak signal-to-noise ratio（PSNR）,contrast-to-noise（CNR）and structure similarity（SSIM）were applied to assess the reconstructed image quality.The results indicated that both the 60°and 90°rotation increments could be used in the real experiments,and which one to choose may depend on the application requirement.Considering the convenience of operation,the 90°rotation increments is a prior choice.Then,based on the previous simulation research,we built up a 90°rotatory dual-head PET prototype system.In this dissertation,a geometric calibration phantom was designed and then used to calibrate the geometric offset of the system.With the geometric calibration,the artifacts in the reconstructed images were greatly eliminated.Then,we measured the imaging performance of the prototype system.Firstly,the home-made point source and Derenzo phantom is used to measure the spatial resolution.Secondly,the sensitivity and image quality is measured.The results show that the spatial resolution can reach to 1.2 mm and sensitivity is about 5%.Thirdly,the in vivo mouse experiment was carried out and further demonstrated the potential of our system in small animal studies.The last but not the least,when the seperation between the two detector heads was fixed,the MC-SRM will be suitable as the MC-SRM is usefull to improve the quality of reconstructed images.Nevertheless,if there is a need to adjust the separation,for example,the detection of breast,the limitation of MC-SRM will come out because it is time-consumed.For this problem,a SRM construction scheme based on distance-driven（DD）and solid-angle was provided,and we called it as DDSolidA scheme.In this scheme,the calculation time was reduced drastically.The sub-sampling schemes based on voxels,including the sub-voxel scheme and multi-DD scheme,were introduced to improve the quality of the model.In order to verify the effect of using the SRMs based on DDSolidA scheme in the reconstruction,the spatial resolution,noise level were applied to analyze the reconstructed results and then compared with that based on the MC simulation and Multi-ray model.The results indicate that the rods with 1.0 mm diameter are resolved when the voxel is virtually divided into 2×2×2 and 4×4×4 sub-voxels,which is comparable to that with the MC-based SRM.When the voxel is divided into 4 cuboid-shaped sub-voxels,the rods with 1.0 mm diameter are resolved basically with the 80^th iteration was used.