Liang Xinglong, Han Luyi, Zhang Xinlin, Li Xinnian, Sun Yue, Tong Tong, Tan Tao, Mann Ritse
The Department of Radiology and Nuclear Medicine, Radboud University Medical Centre, Nijmegen, The Netherlands.
The Department of Radiology, The Netherlands Cancer Institute, Amsterdam, The Netherlands.
Med Phys. 2025 Jul;52(7):e17860. doi: 10.1002/mp.17860. Epub 2025 Apr 28.
Magnetic resonance imaging (MRI) is a crucial medical imaging technique that can determine the structural and functional status of body tissues and organs. However, the prolonged MRI acquisition time increases the scanning cost and limits its use in less developed areas.
The objective of this study is to design a lightweight, data-driven under-sampling pattern for fastMRI to achieve a balance between MRI reconstruction quality and sampling time while also being able to be integrated with deep learning to further improve reconstruction quality.
In this study, we attempted to establish a connection between k-space and the corresponding MRI through singular value decomposition(SVD). Specifically, we apply SVD to MRI to decouple it into multiple components, which are sorted by energy contribution. Then, the sampling points that match the energy contribution in the k-space, which correspond to each component are selected sequentially. Finally, the sampling points obtained from all components are merged to obtain a mask. This mask can be used directly as a sampler or integrated into deep learning as an initial or fixed sampling points.
The experiments were conducted on two public datasets, and the results demonstrate that when the mask generated based on our method is directly used as the sampler, the MRI reconstruction quality surpasses that of state-of-the-art heuristic samplers. In addition, when integrated into the deep learning models, the models converge faster and the sampler performance is significantly improved.
The proposed lightweight data-driven sampling approach avoids time-consuming parameter tuning and the establishment of complex mathematical models, achieving a balance between reconstruction quality and sampling time.
磁共振成像（MRI）是一种关键的医学成像技术，能够确定身体组织和器官的结构与功能状态。然而，MRI采集时间过长会增加扫描成本，并限制其在欠发达地区的应用。
本研究的目的是为快速MRI设计一种轻量级、数据驱动的欠采样模式，以在MRI重建质量和采样时间之间取得平衡，同时还能够与深度学习相结合，进一步提高重建质量。
在本研究中，我们试图通过奇异值分解（SVD）在k空间与相应的MRI之间建立联系。具体而言，我们将SVD应用于MRI，将其解耦为多个分量，并按能量贡献进行排序。然后，依次选择k空间中与每个分量对应的能量贡献相匹配的采样点。最后，将从所有分量获得的采样点合并以获得一个掩码。这个掩码可以直接用作采样器，或者作为初始或固定采样点集成到深度学习中。
在两个公共数据集上进行了实验结果表明，当基于我们的方法生成的掩码直接用作采样器时，MRI重建质量超过了现有最先进的启发式采样器。此外，当集成到深度学习模型中时，模型收敛更快，采样器性能显著提高。
所提出的轻量级数据驱动采样方法避免了耗时的参数调整和复杂数学模型的建立，在重建质量和采样时间之间取得了平衡。
