SPECT图像处理与重建的压缩感知优化方法研究

本项目聚焦于SPECT图像处理与重建的压缩感知优化方法研究。传统的解析重建算法适合完备的投影数据，不满足医学成像降低辐射剂量的要求；而迭代法适合投影数据不完备但无法满足重建精度和速度的要求，因此研究稀疏投影数据条件下的SPECT图像重建，具有重要的价值。将压缩感知理论(CS)引入到SPECT图像重建与ML-EM、OS-EM等算法研究中，拟首先建立它们间内在的相互联系与转化，有效解决图像的稀疏基问题；拟利用观测矩阵与稀疏基的非相关准则研究对应的观测矩阵设计问题，以最大非相关性为目标，构建获取观测矩阵的极值问题及相应的算法；拟有效解决基于SPECT冗余稀疏字典下图像重建问题，研究得到SPECT图像复杂度低、观测维数少、稳定的高精度重建方法。本项目将进一步完善和丰富压缩感知理论的研究途径，拓展其应用范围，进一步推动SPECT图像重建优化方法的研究。

This project will focus on compressed sensing optimal methods research in SPECT image processing and reconstruction. The traditional analytical reconstruction algorithms are not able to meet the requirements of medical imaging to reduce the radiation dose under complete projection data, and the traditional iterative method are not able to meet the requirements of the reconstruction accuracy and speed of medical diagnosis under incomplete projection data. Therefore, it is very important to study SPECT image reconstruction under sparse projection data. The theory of compressed sensing is introduced into the study of SPECT image reconstruction, ML-EM, OSEM and other algorithms. This project will be proposed to first establish the intrinsic interconnections and transformation between them, and effectively solve the problem of the sparse base representation of image; It will be also proposed to use the incoherence criteria of observation matrix and sparse base to study the design problem of observation matrix corresponding to sparse base, and to build the extreme value problem and corresponding algorithms for obtaining observation matrix with the goal of maximum incoherence; under compressed sensing, the problem of SPECT image reconstruction based on SPECT redundant sparse dictionaries will intend to be effectively solved, and high-precision reconstruction methods with low complexity, few observation measurements, and stability in SPECT images will be studied. In this project the research methods of compressed sensing theory itself will be further improved and enriched, its application range will be expanded, and the optimization methods research of SPECT image reconstruction will be further promoted.