results: 本研究提出了一种扩展的Spectro-Dynamic MRI框架，可以重建1) 时间分辨率为11 ms的MR图像，2) 时间分辨率为11 ms的运动场图像，3) 动态参数，以及4) 活动力的激活力。此外，该方法还比two-step方法（先从不含运动信息的下探测数据中重建时间分辨率MR图像，然后使用运动场图像进行运动估算）表现更好。Abstract
Measuring the dynamics and mechanical properties of muscles and joints is important to understand the (patho)physiology of muscles. However, acquiring dynamic time-resolved MRI data is challenging. We have previously developed Spectro-Dynamic MRI which allows the characterization of dynamical systems at a high spatial and temporal resolution directly from k-space data. This work presents an extended Spectro-Dynamic MRI framework that reconstructs 1) time-resolved MR images, 2) time-resolved motion fields, 3) dynamical parameters, and 4) an activation force, at a temporal resolution of 11 ms. An iterative algorithm solves a minimization problem containing four terms: a motion model relating the motion to the fully-sampled k-space data, a dynamical model describing the expected type of dynamics, a data consistency term describing the undersampling pattern, and finally a regularization term for the activation force. We acquired MRI data using a dynamic motion phantom programmed to move like an actively driven linear elastic system, from which all dynamic variables could be accurately reconstructed, regardless of the sampling pattern. The proposed method performed better than a two-step approach, where time-resolved images were first reconstructed from the undersampled data without any information about the motion, followed by a motion estimation step.
摘要
An iterative algorithm solves a minimization problem with four terms: a motion model relating motion to fully-sampled k-space data, a dynamical model describing expected dynamics, a data consistency term describing undersampling pattern, and a regularization term for activation force. We acquired MRI data using a dynamic motion phantom programmed to move like an actively driven linear elastic system, from which all dynamic variables could be accurately reconstructed, regardless of sampling pattern. The proposed method outperformed a two-step approach where time-resolved images were first reconstructed without motion information and then motion was estimated.