融合联合全变分和WSURE参数选择的并行磁共振成像重建OA
Parallel magnetic resonance imaging reconstruction via integration of joint total variation and WSURE parameter selection
基于自一致性并行成像技术(SPIRiT)的算法面临重建速度慢、内存消耗大以及参数选择困难的问题,为此,采用算子分裂技术将基于联合全变分的SPIRiT模型(JTV-SPIRiT)分解为梯度计算子问题和可通过快速梯度投影(FGP)方法求解的JTV去噪子问题,以简化参数选择;引入基于蒙特卡洛的WSURE(weighted stein's unbiased risk estimate)方法,实现JTV-SPIRiT复合模型的参数自动选择.实验结果表明,提出的方法在关键图像质量指标上与基准方法相当,同时,其重建速度获得了显著提升,且利用欠采样数据即可实现近似最优的参数选择.
Existing self-consistent parallel imaging(SPIRiT)algorithms face challenges in reconstruction speed,memory consumption,and parameter selection.To address these,this study employs an operator splitting technique to decompose the Joint Total Variation-based SPIRiT model(JTV-SPIRiT)into two subproblems:a gradient computation part and a JTV denoising part solved efficiently via the Fast Gradient Projection(FGP)method.This approach not only accelerates computation but also simplifies parameter selection.Moreover,a Monte Carlo-based weighted Stein's unbiased risk estimate(WSURE)method is introduced to enable fully automatic parameter selection.Experimental results demonstrate that the proposed method achieves comparable performance to baseline approaches in key image quality metrics,while significantly improving reconstruction speed.Moreover,near-optimal parameter selection can be achieved using only undersampled data.
何泽鑫;段继忠
昆明理工大学 信息工程与自动化学院,昆明 650504昆明理工大学 信息工程与自动化学院,昆明 650504
信息技术与安全科学
磁共振成像图像重建联合全变分参数选择基于蒙特卡洛的WSURE
magnetic resonance imagingimage reconstructionjoint total variationparameter selectionMonte Carlo-based weighted Stein's unbiased risk estimate(WSURE)
《重庆邮电大学学报（自然科学版）》 2026 (2)
310-320,11
云南省基础研究计划项目面上项目(202301AT070452) Yunnan Fundamental Research Project(202301AT070452)
评论