Lung Navigation System for Localizing and Resecting Nodules

项目摘要 虽然楔形切除术可以改善肺功能，但肿瘤复发率几乎为 是肺叶切除术的两倍，5年生存率明显更低。这可能是由于困难。 准确定位和切除肺内的结节。目前，没有有效的方法， 精确定位结节并将外科缝合装置引导至最佳切除边缘。的 这项研究的长期目标是研究管理肺结节的算法和技术， 诊断到手术切除。本提案的目的是设计和开发肺导航 （LungNav）系统定位和切除，具有足够的边缘，小和早期恶性肺结节。的 实验方法将是设计和开发与活动结节集成的LungNav系统 跟踪器称为J-bar，机器学习算法用于确定最佳切除边缘，并跟踪 用于精确切除结节的外科缝合装置。肿瘤变形算法和增强 将开发现实显示器，以便在胸腔镜视频中可视化结节，并引导外科吻合器 实时调整到最佳边缘。假设通过将J形杆锚定在结节附近，结节 尽管当肺萎陷时组织显著变形， 在手术中被操纵为了实现这一项目的目标，我们将努力实现以下具体目标： 1)设计和开发结节跟踪器（J-bar）和变形算法，以估计实时位置 的结节。2)研究基于卷积神经网络（CNN）的机器学习方法， 确定最佳切除边缘。3)设计和开发一个软件导航模块，称为LungNav， 用于使肿瘤可视化并将外科缝合器导航到最佳切除边缘。4)验证 使用离体肺组织和活猪模型的LungNav系统的设计和性能。的 拟议的研究是重要的，因为它解决了一个重要的问题，这可能会影响几个 每年有1000名患者，准确定位和切除肺结节，同时保留健康的肺 功能这项研究是创新的，因为它建立在最先进的机器学习算法，导航， 系统和增强现实方法来准确地诊断和定位结节， 明显的组织变形。该项目的预期成果是开发基于CNN的机器 用于肺结节分类的学习算法和具有肿瘤变形算法的LungNav系统，以及 增强现实方法来定位和引导肺结节的完整手术切除。

Project Abstract Although Wedge Resection Surgery results in better lung function, tumor recurrence rate is almost double that of lobectomy, with significantly poorer 5-year survival rates. This may be attributed to the difficulty in accurately localizing and resecting the nodules in a deflated lung. Currently, there is no effective method of accurately localizing the nodules and guiding the surgical stapling device to the optimal resection margin. The long-term goal of this research is to investigate algorithms and technologies to manage lung nodules from diagnosis to surgical resection. The objective of this proposal is to design and develop a lung navigation (LungNav) system to localize and excise, with sufficient margin, small and early malignant lung nodules. The experimental methods will be to design and develop the LungNav system integrated with an active nodule tracker called J-bar, machine learning algorithms for determining the optimal resection margin, and tracked surgical stapling device for accurately excising the nodule. Tumor deformation algorithms and augmented reality displays will be developed to visualize the nodule on thoracosopy videos and guide the surgical stapler in real-time to the optimal margin. The hypothesis is that by anchoring the J-bar close to the nodule, the nodule position can be accurately tracked in real-time despite significant tissue deformation when the lung is collapsed and manipulated during surgery. To achieve the goals of this project, we will pursue the following specific aims: 1) Design and develop the nodule tracker (J-bar) and deformation algorithms to estimate the real-time position of the nodule. 2) Investigate a machine-learning approach based on convolutional neural networks (CNN) to determine the optimal resection margin. 3) Design and develop a software navigation module, called LungNav, for visualizing the tumor and navigating the surgical stapler to the optimal resection margin. 4) Validate the design and performance of the LungNav system using ex-vivo lung tissue and live porcine models. The proposed research is significant since it addresses an important problem, which potentially affects several thousand patients each year, of accurately localizing and resecting lung nodules while preserving healthy lung function. The research is innovative since it builds on state-of-the-art machine learning algorithms, navigation systems and augmented reality methods to accurately diagnose and localize the nodule in presence of significant tissue deformation. The expected outcome of the project is the development of CNN-based machine learning algorithms for lung nodule classification and a LungNav system with tumor deformation algorithms and augmented reality methods to localize and guide complete surgical resection of lung nodules.