Robust Compressed Sensing MRI with Deep Generative Priors
Robust Compressed Sensing MRI with Deep Generative Priors
复制标题
DOI：
--
复制
发表时间：
2021-08
期刊：
影响因子：
--
通讯作者：
A. Jalal;Marius Arvinte;Giannis Daras;E. Price;A. Dimakis;Jonathan I. Tamir
中科院分区：
文献类型：
--
作者：
A. Jalal;Marius Arvinte;Giannis Daras;E. Price;A. Dimakis;Jonathan I. Tamir
CSGM框架（Bora-Jalal-Price-dimakis'17）表明，深层生成先验可以是解决反问题的强大工具。但是，迄今为止，该框架在某些数据集（例如人的面孔和MNIST数字）上仅在经验上取得了成功，并且众所周知，在分布样本中的性能差。在本文中，我们介绍了CSGM框架在临床MRI数据上的首次成功应用。我们对FastMRI数据集进行了脑部扫描的生成培训，并证明通过Langevin Dynamics进行后验采样可实现高质量的重建。此外，我们的实验和理论表明，后采样对地面真实分布和测量过程的变化是鲁棒的。我们的代码和模型可在：\ url {https://github.com/utcsilab/csgm-mri-langevin}中获得。
The CSGM framework (Bora-Jalal-Price-Dimakis'17) has shown that deep generative priors can be powerful tools for solving inverse problems. However, to date this framework has been empirically successful only on certain datasets (for example, human faces and MNIST digits), and it is known to perform poorly on out-of-distribution samples. In this paper, we present the first successful application of the CSGM framework on clinical MRI data. We train a generative prior on brain scans from the fastMRI dataset, and show that posterior sampling via Langevin dynamics achieves high quality reconstructions. Furthermore, our experiments and theory show that posterior sampling is robust to changes in the ground-truth distribution and measurement process. Our code and models are available at: \url{https://github.com/utcsilab/csgm-mri-langevin}.