数字乳腺断层摄影(DBT)深度分辨率极值建模及逼近极值重建

数字乳腺断层摄影（DBT）是乳腺癌早期筛查的新技术，它利用小角度范围内的有限次投影，重建乳腺准三维图像。由于扫描方式限制，DBT图像深度分辨率低于平面分辨率10倍以上，导致处于断层之间远离断层焦平面的微小病灶无法清晰成像，易造成漏诊。提高深度分辨率是解决该问题的有效途径。现有研究仅给出了深度分辨率与少数成像参数之间的定性关系，相互之间的定量关系尚未见报道，致使深度分辨率提高策略缺乏详实的理论依据。本课题针对DBT成像技术，对周期性体模进行连续射线积分与滤波反投影，从理论上构建深度分辨率与成像参数之间的关系模型，揭示深度分辨率的极值分布；在此基础上，设计多尺寸重建网格及自适应扩散滤波正则化重建算法，实现深度分辨率逼近极值；借助分布式计算框架，提出一种适用于DBT重建的云计算加速方案，解决重建耗时问题。本研究有望为DBT系统结构优化提供理论依据，为DBT增强对乳腺癌微小病灶检测提供解决方案。

Digital breast tomosynthesis (DBT) is a new technology for early screening of breast cancer, it uses a limited number of projections over a small angle range to reconstruct a pseudo-3D image of the breast. Due to the limitation of the scanning method, the depth resolution of the DBT image is less than 10 times the plane resolution, which results in the inability of the micro-lesions located between the slices and far away from the slice focal planes to be clearly imaged, which may result in missed diagnosis. Increasing the depth resolution is an effective way to solve this problem. The existing research only gives the qualitative relationship between the depth resolution and a few imaging parameters, and the quantitative relationship between them has not been reported, resulting in a lack of detailed theoretical basis for the depth resolution improvement strategy. This topic is aimed at DBT imaging technology, continuous ray integration and filtered back-projection is performed on a periodic phantom to theoretically construct a relation model between depth resolution and imaging parameters and to reveal the extremum distribution of the depth resolution. On this basis, a multi-size reconstruction mesh and an adaptive diffusion filter regularization reconstruction algorithm are designed to realize the approaching extremum reconstruction of the depth resolution. With the help of a distributed computing framework, a cloud computing acceleration scheme suitable for DBT reconstruction is proposed to solve the reconstruction time-consuming problem. This study is expected to provide a theoretical basis for the structural optimization of DBT system and provide a solution for DBT to enhance the detection of breast cancer micro-lesions.