面向低剂量乳腺CT的影像链复合优化技术研究

At present, dose level of breast CT imaging is equal to or under mammography, still the effort of decreasing the dose namely reducing the risk of carcinogenicity is employed. If a lower X-ray exposure is used for breast CT scan to reduce the radiation dose of patients, reconstructed breast CT thin sections will be significantly degraded by the quantum noise, inherent noise and X-ray scatter of breast CT system. In recent years, how to remove the noise and scatter in the reconstructed volume to improving the conspicuity of breast lesions while also retaining spatial resolution has been the subject of intense study. This project is to study on composite optimization techniques of imaging chain for low-dose breast CT. In the front end of the imaging chain, this project will develop irregularly shaped filters of X-ray with two-dimensional modulation capability based on properties of breast CT scan and structural characteristics of the breast. In the back end of the imaging chain, this project mainly exploits the statistical properties of noise, spatial local features and temporal correlations of projection data, and aims at developing spatial-temporal denoising algorithms for breast CT. By employing multiple of imaging chain modulation, it is capable of reducing scatter and noise levels, removing image artifacts and enhancing lesion characteristics in reconstructed breast images. Consequently, it is possible to improve the detection rate of breast lesions while also reduce the patient radiation dose significantly, which can develop the effectiveness of breast CT examinations and safeguard patients' security.

尽可能地降低辐射剂量以减小辐射危害是研制乳腺CT的关键问题。由于乳腺CT本身存在一定的系统噪声和射线散射，采用低剂量采集又会引起量子噪声的大幅增加，使得重建图像退化严重，难以用于临床诊断。近年来，有效抑制低剂量CT图像中噪声和散射干扰，并保持图像空间分辨率的去噪技术吸引着越来越多研究者的关注。本项目面向低剂量乳腺CT的影响链进行复合优化研究。在影像链的前端，根据乳腺CT的数据采集方式及乳腺受体的结构特点，设计对X射束具有二维调制能力的异形过滤器。在影像链的后端，利用投影数据的噪声统计特性、空间局部特性和时间相关性，建立时空结合的投影域三维去噪算法。通过对乳腺CT影像链进行多重调制，有效降低噪声水平、减小散射强度、抑制图像伪影并增强乳腺图像中钙化点和肿瘤等病灶特征，能够在降低扫描辐射剂量下，保持或提高恶性病变的检出率，充分发挥乳腺CT检查的有效性，并保障受检人群的安全性。

乳腺CT作为一种临床医用设备，在保证图像质量的前提下，尽可能地降低辐射剂量一直是研制乳腺CT需要考虑的重要问题。本项目通过研究乳腺CT影像链的复合优化技术，以解决乳腺CT中存在的辐射剂量、射线散射、射束硬化和噪声退化等诸多问题。在影像链的前端，基于蒙特卡洛模拟及成像实验的方法，研究并设计了对X光谱具有空间调制能力的异形过滤器。实验结果表明异形过滤器的剂量效率可以达到常规平板过滤器的四倍以上，能够降低由射线散射和射束硬化造成的图像伪影，实现了对辐射剂量的控制并提高图像质量。在影像链的后端，针对传统各向异性扩散模型的病态性及扩散系数的低适应性，提出了基于归一权梯度函数投影域去噪模型及基于时变扩散系数的投影域去噪模型。实验结果表明该方法能够提高图像对噪比到三倍以上，在保证算法效率的同时，有效地实现噪声的抑制和图像细节特征的保留。通过项目研究，有效地提高了乳腺三维成像的剂量效率，增强了乳腺图像中钙化点和肿瘤等病灶特征，对于保证病人辐射安全，提高临床诊断率具有重要意义。基于本项目的研究成果，共申请发明专利两项，项目执行期间均已获得授权。发表录用论文四篇，并培养了多名在相关领域从事科研工作的青年骨干和研究生。