Ultrasound based neurosurgical navigation with uncertainty visualization

手术切除是几乎所有脑肿瘤的初始治疗方法，手术切除的程度与 预后。然而，因为脑瘤，特别是胶质瘤，与周围功能正常的大脑密切相关 对于组织，积极的切除必须在造成新的神经缺陷的风险之间取得平衡。现代科技的进展 解剖和功能成像以及神经导航的广泛采用现在帮助神经外科医生计划和 采用最佳手术方式。不幸的是，手术期间大脑的形状发生了变化，这就是所谓的大脑 Shift，推翻了所有商业神经导航系统可以映射到术前数据的假设 使用刚性配准的患者坐标。因为脑部转移在手术过程中进行，神经的严格注册- 当切除边缘组织时，导航系统在切除的关键最后阶段最不准确。 已经有20多年的研究投资于测量、建模和补偿大脑转移 为神经导航系统提供从术前图像数据到 病人的大脑出现脑部移位。虽然结果很有希望，但它们还不够准确，不足以 整合到商业系统中。非刚性配准受测量和建模不确定性的影响 在整个3D空间中，这一点各不相同。大多数非刚性配准方法不会试图量化这种不确定性， 据我们所知，没有人试图将这种不确定性呈现给外科医生。我们认为，这一点很重要 让外科医生意识到这种不确定性，以便他们能够做出明智的决定，特别是在 不确定性很高。在这个项目中，我们计划研究模拟注册的非刚性注册算法 利用配准的显式不确定性、半自动和全自动非刚性配准方法 迭代指导注册改进的不确定性，以及有效呈现的可视化范例 外科医生在外科环境中注册的不确定性。 我们假设，脑移位矫正的配准不确定性的有效表示和可视化 神经导航将导致迭代半自动和全自动非刚性配准方法，从而改进 配准准确性和2)允许神经外科医生在肿瘤切除过程中做出更明智的决定，这将导致 以提高影像引导神经外科手术的临床效果。我们将实现以下目标：1.开发新颖的特色- 基于显式表示不确定性的图像配准算法；2.使用配准不确定性图指导 半自动和全自动非刚体配准；3.用不确定性可视化评价非刚体配准的实用性 在临床环境中。

Surgical resection is the initial treatment for nearly all brain tumors and the extent of resection is strongly correlated with prognosis. However, because brain tumors, especially gliomas, are intimately involved in surrounding functioning brain tissue, aggressive resection must be balanced against the risk of causing new neurological deficits. Modern advances in anatomical and functional imaging and the widespread adoption of neuro-navigation now help neurosurgeons to plan and execute an optimal surgical approach. Unfortunately, changes in the shape of the brain during surgery, known as brain shift, invalidate the assumption of all commercial neuro-navigation systems that preoperative data can be mapped to patient coordinates using rigid registration. Because brain shift progresses during surgery, the rigid registration of neuro- navigation systems is least accurate at the critical final stages of resection when the marginal tissue is being removed. There has been more than 20 years of research invested in measuring, modeling and compensating for brain shift with the goal of providing neuro-navigation systems with an accurate nonrigid registration from preoperative image data to the patient’s brain in the presence of brain shift. While results are promising, they are not yet accurate enough to be incorporated into commercial systems. Nonrigid registration is subject to both measurement and modeling uncertainty that varies throughout 3D space. Most nonrigid registration methods do not attempt to quantify this uncertainty and, to our knowledge, there have been no attempts to present this uncertainty to the surgeon. We believe that it is important to make surgeons aware of this uncertainty so that they can make informed decisions, particularly in locations where uncertainty is high. In this project, we plan to investigate nonrigid registration algorithms that model registration uncertainty explicitly, semi-automatic and fully-automatic nonrigid registration methods that utilize registration uncertainty to iteratively guide registration improvements, and visualization paradigms for effective presentation of registration uncertainty to surgeons in the surgical environment. We hypothesize that effective representation and visualization of registration uncertainty for brain shift correction in neuro-navigation will 1) lead to iterative semi-automatic and fully-automatic nonrigid registration methods that improve registration accuracy and 2) allow neurosurgeons to make more informed decisions during tumor resections that will lead to increased clinical impact of image-guided neurosurgery. We will carry out the following Aims: 1. Develop novel feature- based image registration algorithms that represent uncertainty explicitly; 2. Use registration uncertainty maps to guide semi- and fully-automatic nonrigid registration; 3. Evaluate the utility of nonrigid registration with uncertainty visualization in a clinical setting.