基于噪声相关性的惩罚加权最小二乘算法在低剂量数字乳腺层析成像中的应用

摘要

Objective To achieve low-dose digital breast tomosynthsis (DBT) projection recovery using penalized weighted least square algorithm incorporating accurate modeling of the variance of the projection data and noise correlation in the flat panel detector. Methods Models were established for the quantal noise and electronic noise in the DBT system to construct the penalized weighted least squares algorithm based on noise correlation for projection data restoration. The filter back projection algorithm was then used for DBT image reconstruction. Results The reconstruction results of the ACR phantom data at different dose levels showed a good performance of the proposed method in noise suppression and detail preservation. CNRs and LSNRs of the reconstructed images from the restored projections were increased by about 3.6 times compared to those of reconstructed images from the original projections. Conclusion The proposed method can significantly reduce noise and improve the quality of DBT images.%目的 对投影数据方差的精确建模并结合DBT平板探测系统的噪声相关性构建精准噪声模型下的基于噪声相关性的惩罚加权最小二乘算法在低剂量乳腺层析成像图像中的应用.方法 首先对投影数据进行量子噪声和电子噪声建模,使以往常用的近似噪声模型精准化,然后构建基于噪声相关性的惩罚加权最小二乘算法用于投影数据恢复;最后对处理后的投影数据采用滤波反投影算法进行重建.结果 对不同剂量下ACR标准体模数据进行处理得到的重建结果噪声明显降低,细节对比度提高.恢复投影数据的重建图像与原始数据重建图像相比,CNRs和LSNRs提升了约3.6倍.结论 对投影数据噪声抑制效果明显,重建DBT图像质量有很大的提升.