基于双字典自适应学习算法的低采样率CT重建  

Low Sampling Rate CT Reconstruction Based on Dual Dictionary Adaptive Learning Algorithm

作　　者：栾峰 杨帆[1,2] 蔡睿智 杨晨[1,2] LUAN Feng;YANG Fan;CAI Rui-zhi;YANG Chen(School of Computer Science&Engineering,Northeastern University,Shenyang 110169,China;Key Laboratory of Intelligent Computing in Medical Image of Ministry of Education,Northeastern University,Shenyang 110169,China)

机构地区：[1]东北大学计算机科学与工程学院,辽宁沈阳110169 [2]东北大学医学影像智能计算教育部重点实验室,辽宁沈阳110169

出　　处：《东北大学学报（自然科学版）》2022年第12期1709-1716,共8页Journal of Northeastern University(Natural Science)

摘　　要：在医疗诊断中,稀疏采样能减少CT扫描过程中辐射对患者的伤害.但直接对稀疏采样后的投影数据进行重建,会使CT重建后的图像出现失真、伪影等问题.为保证低采样率下重建图像的质量,提出了双字典自适应学习算法,参照Sparse-Land模型的双字典学习框架,将K-SVD算法与双字典学习算法框架相结合得到补全投影数据,利用FBP算法进行重建得到高质量的重建图像.实验结果表明,在低采样率下使用所提方法进行CT重建的图像质量优于COMP双字典学习算法和MOD双字典学习算法,并且此方法有效提高了CT图像重建在低采样率时的性能.In medical diagnosis, sparse sampling can reduce radiation damage to patients during CT scanning. However, direct reconstruction of sparse sampling projection data will cause distortion and artifacts in the reconstructed CT images. In order to ensure the quality of reconstructed images at low sampling rate, a dual dictionary adaptive learning algorithm is proposed, referring to the dual dictionary learning framework under the Sparse-Land model. K-SVD algorithm is combined with the dual dictionary learning algorithm framework to obtain patched projection data and FBP(filter back projection) algorithm is used to reconstruct high-quality reconstructed images. Experimental results show that the proposed method is superior to COMP double dictionary learning algorithm and MOD double dictionary learning algorithm in CT reconstruction at low sampling rate, and this method effectively improves the performance of CT image reconstruction at low sampling rate.

关 键 词：CT图像重建 K-SVD算法 双字典学习算法 自适应学习算法 FBP算法 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]