Image quality improvement of microPET images by an accelerated reconstruction algorithm and filters

• Main Authors: Chun-Cheng Lin, 林駿丞
• Other Authors: Jyh-Cheng Chen
• Format: Others
• Language: zh-TW
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/36340911317664951757

碩士 === 國立陽明大學 === 放射醫學科學研究所 === 95 === Abstract Introduction: PET can provide in vivo, quantitative and such as maximum likelihood expectation maximization (MLEM) algorithm are rapidly becoming the standard for image reconstruction in emission computed tomography. Because a large number of iterations may be required, MLEM algorithm has very slow convergent speed. The objective of our study was to develop a fast and better reconstruction algorithm and improve the image quality of the reconstruction algorithm by implementing filter. Methods and Materials: We developed a new algorithm called modified ordered subset expectation maximization (MOSEM) algorithm. In the algorithm, the number of projections in each subset among different iterations is variable to achieve fast convergence.We further implemented and evaluated an advanced version of the MOSEM for microPET image reconstruction, inter-update Metz filtered MOSEM (IMF-MOSEM). The IMF-MOSEM incorporates filtering action into the image updating process to improve the quality of the reconstruction. This study used Metz filter to improve image quality. Metz filter is an enhanced image filter. It is modified from Gaussian filter and changes Metz power to achieve local enhancement. Our study used four kinds of image processing, including post-processing, pre-processing, IMF-OSEM and IMF-MOSEM. In order to achieve accelerated reconstruction algorithm and improve image quality. At first, we used digital rat brain to evaluate the speed of MOSEM. Second, we used concentric circle physical phantom to evaluate the result of different processing. The result was evaluated by CR (contrast recovery), CV (coefficient of variation) and FWHM (Full-Width Half-Maximum). Results: For the same image quality, the reconstruction time of traditional OSEM was 20.51 seconds and the MOSEM was 35.16 seconds in two dimension digital rat brain study. In physical phantom, the CRs of ordinary, post-processing, IMF-OSEM and IMF-MOSEM image were 5.89, 6.93, 7.36 and 8.55, respectively. The CVs of ordinary, post-processing, IMF-OSEM and IMF-MOSEM image were 0.6, 0.66, 0.86 and 0.87, respectively. The FWHM of ordinary, post-processing, IMF-OSEM and IMF-MOSEM were 1.71, 1.65, 1.63 and 1.54 mm, respectively. Conclusion: In two dimensions digital rat brain, the reconstruction speed of the MOSEM was faster than traditional OSEM by about 41.6%. In IMF-MOSEM, our results depended upon the choosing of Metz filter parameters. When we used the appropriate Metz filter, the CV value was not different form ordinary image but CR value improved by 45.1%. FWHM value was also decreased by 9%. The study demonstrated that our algorithm can be not only used for microPET system but also used to improve the image quality of the clinical PET system. Keyword: MLEM, OSEM, Metz filter, Wiener filter