Real-time cardiovascular MRI and co-registration technical development

在过去的一年里，我们继续增强成像平台，以指导基于心血管导管的治疗。其中包括与传统X射线联合注册的MRI，以及独立的实时MRI。 从MRI数据集获得的静态3D路线图被用于增强X射线心血管介入手术的图像指导，并确实已在该实验室用于开发新的治疗方法，如二尖瓣环扎成形术。静态路线图不能准确反映心脏和呼吸运动期间的心血管解剖结构。我们已经开发了一种系统，可以从实时MRI扫描中测量呼吸和心脏运动，并推导出一套仿射模型，可以用来敲击和呼吸覆盖在实时X光上的3D路线图。我们还开发了健壮的全自动数学技术来配准成像模式之间的基准标记。我们将继续改进该系统的工作流程和可用性，使其能够在常规基础上用于指导程序。 我们开发了一种自适应噪声消除系统，以滤除MRI套件中射频和磁梯度场的干扰，并展示了检测在MR扫描过程中原本被掩盖的心律失常的能力。我们计划将该系统扩展到全多导联表面和心内电信号，并将其集成到临床血流动力学记录系统中。 我们使用面向游戏的图形处理单元提供的廉价并行计算资源来加速计算密集型MRI数据的重建。我们已经成功地将非笛卡尔并行成像集成到交互式采集和重建设置中，并证明了对于相对复杂的重建算法，实时重建和可视化是可能的。这将很快与扫描仪软件集成，以实现与其他序列组件的无缝结合。 我们开发了一个系统，为操作员提供实时MRI数据的多种同时表示，以交互方式平衡时间和空间分辨率。操作员选择所需的表示。我们实现了时间分辨率交互式选择的黄金角度实时磁共振成像。 我们开发了使用最少(高度欠采样投影成像)数据跟踪复杂三维曲线MRI导管设备的高效技术，允许无缝集成到实时MRI可视化系统中。这允许在复杂的导管介入操作中结合二维和三维数据。同样，我们已经开发了一些技术来消除在对有源设备使用加速、多层、并行成像时出现的运动相关重影伪影。 我们正在将我们非常成功的本地实时磁共振软件环境迁移到一个商业平台上，以促进NIH以外的翻译，并加强产业界和大学的合作。这需要相当大的发展来更新主要的实时MRI脉冲序列，以促进快速的多作者或多机构原型制作。 我们还继续开发实时磁共振成像的新方法，或设计实时磁共振成像的本地非商业性实施例，以适应正在开发的程序的需要。例如，我们正在开发具有螺旋轨迹的实时交互流动成像。我们开发了一种新的最优在线梯度波形设计，用于快速流动不敏感或流动敏感成像。这需要开发新的梯度波形代数并集成到函数序列中。我们还在研究几种脉冲序列方法，以减少实时核磁共振期间导电设备的入射加热。最后，我们正在开发定制的用户输入设备，以帮助导管操作员更有效地开出所需的切片处方，而不会分散他们对主要导管操作程序的注意力。

During the past year we have continued enhancement of imaging platforms to guide cardiovascular catheter based treatments. These have included co-registered MRI with conventional X-ray, as well as standalone real-time MRI. Static 3D roadmaps derived from MRI datasets are used to enhance image guidance for X-ray cardiovascular interventional procedures, and indeed have been used in this lab to develop novel treatments such as mitral cerclage annuloplasty. Static roadmaps do not accurately represent cardiovascular anatomy during cardiac and respiratory motion. We have developed a system to measure respiratory and cardiac motion from real-time MRI scans and to derive a set of affine models which can be used to beat and breath the 3D roadmaps overlaid on live X-ray. We also have developed robust fully automatic mathematical techniques to register fiducial markers between imaging modalities. We will continue to improve the workflow and usability of the system so that it can be used on a routine basis to guide procedures. We have developed a system of adaptive noise cancellation to filter out interference from the radiofrequency and magnetic gradient fields in an MRI suite, and demonstrated the ability to detect arrhythmia otherwise obscured during MR scanning. We plan to expand the system to full multiple lead surface and intracardiac electrograms, and to integrate this into a clinical hemodynamic recording system We have used inexpensive parallel computing resources afforded by game-oriented Graphics Processing Units to accelerate reconstruction of computationally-intensive MRI data. We have successfully integrated non-Cartesian parallel imaging in an interactive acquisition and reconstruction setup and demonstrated that real-time reconstruction and visualization is possible for relatively complicated reconstruction algorithms. This will soon be integrated with the scanner software to allow seamless combination with other sequence components. We have developed a system to provide the operator multiple simultaneous representations of real-time MRI data balancing temporal and spatial resolution interactively. The operator chooses the desired representation. We have implemented golden-angle real-time MRI with interactive selection of the temporal resolution. We have developed highly efficient techniques to track complex 3 dimensional curve MRI catheter devices using minimal (highly undersampled projection imaging) data, allowing for seamless incorporation into a real-time MRI visualization system. This allows combination of two- and three-dimensional data in complex catheter interventional procedures. Similarly, we have developed techniques to eliminate motion-related ghosting artifact that appears when accelerated, multi-slice, parallel imaging is used for active devices. We are migrating our highly successful local real-time MRI software environment onto a commercial platform to facilitate translation outside of NIH, and to enhance industry and university collaboration. This has required considerable development to update workhorse real-time MRI pulse sequences to facilitate rapid multi-author or multi-institution prototyping. We also continue to develop new approaches to real-time MRI, or to engineer local noncommercial embodiments of real-time MRI to suit the needs of procedures being developed. For example we are developing real-time interactive flow imaging with spiral trajectories. We have developed a new optimal on-line gradient waveform design for rapid flow-insensitive or flow-sensitive imaging. This required development of new gradient waveform algebra and integrated into functional sequences. We are also working on several pulse-sequence approaches to reduce incident heating of conductive devices during real-time MRI. Finally we are developing custom user-input devices to assist catheter operators more effectively to prescribe desired slice prescriptions without distracting from their primary catheter manipulation procedures.