Cone-Beam Methods for Dynamic Volumetric X-Ray CT

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): Currently, medical CT scanners are under rapid development with an increasingly larger cone angle, while biomedical micro-CT scanners are already in cone-beam geometry. Despite the importance of cone-beam CT, cone-beam image reconstruction algorithms are not fully developed. There is a critical and immediate need for a dynamic volumetric performance of cone-beam CT, subject to multiple constraints such as dose, noise, range, contrast, etc. The overall goal of this project is to develop and optimize analytic cone-beam algorithms with an emphasis on high temporal resolution and short scan range, and directly applicable to major applications such as cardiac imaging, CT fluoroscopy, perfusion studies, CT angiography, oncologic imaging, small animal imaging, as well as PET and SPECT. This project is based on the latest cone-beam CT results, and focuses on both approximate and exact reconstruction in the Feldkamp-type, Grangeat-type and Katsevich-type frameworks respectively. The specific aims are to (1) improve Feldkamp-type algorithms for less than half-scan data by scanning pattern design and weighting scheme optimization; (2) extend Grangeat-type half-scan algorithms for long object reconstruction by correcting cone-beam data, and transform the Radon space based reconstruction into the filtered backprojection format; (3) modify Katsevich-type algorithms for dynamic reconstruction by detection coverage minimization and n-PI geometry-based formulation; and (4) evaluate and validate the proposed cone-beam algorithms in theoretical analysis, numerical simulation and phantom experiments, and demonstrate their feasibility and utilities in mouse and patient studies. On completion, superior and practical cone-beam algorithms will have been systematically developed with excellent image quality for dynamic volumetric CT and micro-CT. These proposed algorithms will have been implemented on a PC cluster. The advantages of the algorithms will have been demonstrated in mouse and patient studies.