INTRAOPERATIVE IMAGE GUIDED NEUROSURGERY DVMT

为了支持图像引导神经外科的目标，我们计划 继续我们在开发新的计算机视觉系统方面的工作： 将医学扫描分割成不同解剖结构的图形模型 结构;将这些模型与患者的位置配准， 手术室;提供可视化工具， 看到病人全身的解剖结构 通过微创开口支持手术进入;轨道 并为外科医生提供 通过观察仪器相对于 分段扫描;并自动调整模型以反映 随着手术的进行而变形和变化。 在细分领域，我们将继续临床化发展 用于自动或半自动构建的相关方法 标准医学扫描的解剖模型。 目标是建成 患者特异性模型，直接突出的结构， 外科医生的兴趣。 在一个相关的子提案中，我们提供了一个 将开发新方法以支持这种方法的详细描述 分割技术 在本分项提案的范围内，我们 打算： 通过耦合组织扩展我们目前在分割方面的工作 分类方法与解剖图集。 这将使我们能够 使用预定义的模型作为引导提取的模板 从扫描图中 评估现有和改进的 分割技术 这将包括基线研究， 将我们的分割结果与放射科专家的分割结果进行比较， 并将包括我们的分割结果的准确性验证 与手术过程中观察到的实际解剖结构相比较。 创建直接支持图像引导手术的自动工具， 神经外科，通过将分割的模型直接 与外科病人的关系，用于指导和 导航 这将包括由外科医生评估 分割模型的各个方面更直接相关， 外科手术

To support the goals of image guided neurosurgery, we plan to continue our work in developing novel computer vision systems that: segment medical scans into graphical models of distinct anatomical structures; register such models with the position of the patient in the operating room; provide visualization tools that let the surgeon see the full panoply of the patient's anatomy while at the same time supporting surgical access through minimally invasive openings; track surgical instruments relative to the patient and provide the surgeon with a visualization of the position of the instrument with respect to the segmented scans; and automatically adjust models to reflect deformations and changes as the surgery proceeds. In the area of segmentation, we will continue to develop clinically relevant methods for automatically or semi-automatically constructing anatomical models from standard medical scans. The goal is to build patient- specific models that directly highlight structures of interest to the surgeon. In a related sub-proposal, we provide a detailed description of novel methods to be developed to support such segmentation techniques. Within the context of this sub-proposal, we intend to: Extend our current work in segmentation by coupling tissue classification methods with anatomical atlases. This will allow us to use predefined models to serve as templates for guiding the extraction of structures from scans Evaluate the accuracy and stability of existing and modified segmentation techniques. This will include baseline studies that compare our segmentation results against those of expert radiologists, and will include verification of the accuracy of our segmented results against actual anatomy observed during the surgical procedure. Create automatic tools that directly support Image Guided Surgery in neurosurgery, by bringing the segmented models into direct relationship with surgical patients, for use in guidance and navigation. This will include evaluation by the surgeon of what aspect of segmented models are more directly relevant and useful in the surgical procedure.