Novel Fast Imaging and Reconstruction Strategies for Dynamic MRI

摘要 该研究项目的目标是开发新的，超快速的方法， 成像应用程序，使更大的临床效用在未来。我们打算实现这一目标 通过结合几种现有的图像重建方法，即并行成像， 非笛卡尔轨迹，以产生新的快速采集方法。我们目前的研究 涉及使用径向轨迹，而不是标准的直线轨迹， 在很短的时间内获得高度加速的数据集。这些数据可以 使用称为GRAPPA的并行成像方法的特殊配方重建， 以重建无误差的图像。使用这种技术，我们已经获得了图像， 时间分辨率为60ms。我们计划将这一概念扩展到具有 更快的数据采集潜力，即螺旋和各向异性视场 轨迹使用这些方法，我们相信将有可能生成图像， 小于40 ms，这将允许采集实时、自由呼吸的心脏图像， 使得心电图门控和屏气对于心脏功能检查是不必要的。为了 使这些重建在临床上可接受的时间范围内成为可能， 在GPU平台上实现，这将使重建时间从几分钟减少到 秒 在项目的独立阶段，将利用GPU平台， 研究MRI数据的不同约束重建方法。除了平行 成像和非笛卡尔采集，这些技术，包括压缩 传感也作为可能的快速成像的新的和重要的类别出现 方法.早期的工作表明，数据减少了20倍，因此时间减少了20倍， 需要一个形象。这些方法的力量是显而易见的，尽管目前还不清楚， 它们在临床环境中是可行的，例如，由于令人难以置信的长计算时间， （有时长达数天）。因此，根据我们在本提案第一阶段的经验， 该项目的独立部分将探索这些受限制的潜力 重建方法，并审查将其与非重建方法相结合的可能性。 笛卡尔并行成像方法是早期发展起来的。快速计算 平台，以GPU实现的形式，将允许这些新颖的图像重建 技术进行了大力测试，铺平了道路，这些方法成为实用的， 广泛的临床应用。

Abstract The goal of this research project is to develop new, ultra-fast methods for dynamic imaging applications to enable greater clinical utility in the future. We intend to meet this goal by combining several existing image reconstruction methods, namely parallel imaging and non-Cartesian trajectories, to generate novel fast acquisition methods. Our current research involves the use of radial trajectories, as opposed to the standard, rectilinear trajectory, to acquire highly accelerated datasets in a very short time. These data can then be reconstructed using a special formulation of a parallel imaging method known as GRAPPA in order to reconstruct error-free images. Using this technique, we have acquired images with a temporal resolution of 60ms. We plan to expand this concept to trajectories which have the potential for even fast data acquisition, namely spiral and anisotropic field-of-view trajectories. Using these methods, we believe that it will be possible to generate images in less than 40ms, which will allow the acquisition real-time, free-breathing cardiac images, making EKG gating and breathholding unnecessary for cardiac function exams. In order to make these reconstructions possible in a clinically acceptable timeframe, they will be implemented on a GPU platform, which will reduce the reconstruction time from minutes to seconds. In the independent phase of the project, the GPU platform will be exploited in order to investigate different constrained reconstruction methods for MRI data. In addition to parallel imaging and non-Cartesian acquisitions, these techniques which include compressed sensing have also emerged as a new and important category of possible fast imaging methods. Early work has demonstrated an up to 20-fold reduction in data, and thus time, needed for an image. The power of these methods is obvious, although it is not yet clear if they will be viable in a clinical setting, due to, for instance, incredibly long computation times (sometimes up to days). Thus based on our experience in the first stage of this proposal, the independent portion of this project will explore the potential of these constrained reconstruction methods and examines the possibility of combining them with the non- Cartesian parallel imaging methods developed in the earlier phase. The rapid computational platform, in the form of the GPU implementations, will allow these novel image reconstruction techniques to be vigorously tested, paving the way for these methods to become practical for widespread clinical use.