A Boundary Element Method for MRI/NIR Tomography and Image-guided Fluorescence

描述（由申请人提供）：本提案中的总体假设是新兴的混合成像系统，将光谱学引入成像需要目前不存在的计算工具。具体来说，边界元方法（BEM）算法可以实现对图像引导的乳腺组织近红外（IG-NIR）断层扫描和小动物荧光断层扫描进行三维（3D）图像重建。该系统智能地利用MRI的解剖结构来引导近红外光谱（NIRS）来提高乳腺癌的诊断；和CT组织结构来引导近红外光谱，从而准确地恢复荧光摄取。IG-NIRS的MR和CT方法传统上都依赖于需要体积离散化（如有限元）的扩散方程的数值模型。在本提案中，我们将使用BEM（只需要表面离散化）求解扩散方程，并将其应用于3D图像重建，假设可以从MRI或CT中先验地获得底层组织边界。特别是，将开发一个计算工具箱，使用MRI图像无缝地为不同的组织层（如脂肪、纤维腺和肿瘤）创建表面网格。工具箱将使用这些网格以及近红外测量来重建总血红蛋白，氧饱和度和水的三维组织血管估计；以及每个组织层中散射体大小和数量密度的细胞估计。开发荧光工具箱，从MicroCT图像中获取表面网格，利用边界元法同时求解一组耦合扩散方程，恢复小动物不同组织器官的三维荧光值。这些工具箱与图形用户界面一起，允许3D图像可视化和NIR， MRI和MicroCT图像并并，将为不同的研究小组使用混合成像技术提供易于使用的边界元素软件。利用正在进行的乳房成像试验的临床数据，我们建议分析50例患者的3D边界元组织估计结果，以探索该技术在组织诊断中的敏感性和特异性措施，以及它在非侵入性研究癌症方面的潜力。在ct荧光环境下对小动物进行体内测量也将通过一个单独的资助项目进行，用于测试BEM分子成像。这种新颖的边界元工具箱以其相对于FEM等体积离散化方法的优势，以及在解决图像引导重建问题方面的计算效率，将为三维光学成像设定标准。这将进一步使用MRI-NIR 3D成像作为日常诊断工具，提供病变组织的非侵入性高分辨率功能表征。计划开发两个版本的工具箱，一个是更先进的，可以通过与ART公司的互动转化为商业版本，同时，将分发一个开放访问版本，允许NIRS领域的新手用户使用BEM工具箱建立新的和不断发展的工具。

DESCRIPTION (provided by applicant): The overall hypothesis in this proposal is emerging hybrid imaging systems that bring spectroscopy into imaging require computational tools which currently do not exist. Specifically, a boundary element method (BEM) algorithm can be implemented to perform three dimensional (3D) image reconstruction for image-guided near infrared (IG-NIR) tomography of breast tissue and fluorescence tomography in small animals. This system intelligently utilizes anatomical structures from MRI to guide NIR spectroscopy (NIRS) to improve diagnosis of breast cancer; and CT tissue structure to guide NIRS allowing accurate recovery of fluorescence uptake. Both MR and CT approaches to IG-NIRS traditionally rely on numerical models to the diffusion equation requiring volume discretization (such as finite element). In this proposal, we will solve the diffusion equation using the BEM (requiring only surface discretization) and apply it for 3D image reconstruction, assuming that the underlying tissue boundaries can be obtained a priori from MRI or CT. In particular, a computational toolbox will be developed that seamlessly creates surface meshes for different tissue layers such as adipose, fibroglandular and tumor, using MRI images. The toolbox will use these grids along with NIR measurements to reconstruct 3D tissue vascular estimates of total hemoglobin, oxygen saturation and water; and cellular estimates of scatterer size and number density in each tissue layer. A fluorescence toolbox will also be developed that obtains surface grids from MicroCT images and solves a set of coupled diffusion equations simultaneously using BEM to recover 3D fluorescence values in different tissue organs of small animals. These toolboxes together with a graphical user interface allowing 3D image visualization and juxtaposition of NIR, MRI and MicroCT images, will provide easy-to-use boundary element software for different research groups utilizing hybrid imaging techniques. Leveraging the clinical data from an ongoing breast imaging trial, we propose to analyze the results from 3D boundary element tissue estimates of 50 patients to explore the sensitivity and specificity measures of this technique for tissue diagnosis as well as its potential to study cancer non-invasively. In-vivo measurements from small animals imaged in a CT-fluorescence setting will also be available through a separate funded project, for testing of BEM molecular imaging. This novel BEM toolbox with its strengths over volume discretization methods such as FEM and its computational efficiency in solving the image-guided reconstruction problem will set the standard for 3D optical imaging. This will further the use of MRI-NIR 3D imaging as an everyday diagnostic tool providing non-invasive high-resolution functional characterization of diseased tissue. Two versions of the toolbox will plan to be developed, one which is more advanced and can be translated into a commercial version through interaction with ART Inc, and at the same time, a open access version will be distributed which allows novice users in the field of NIRS to set up new and evolving tools which use the BEM toolbox. PUBLIC HEALTH RELEVANCE: A hybrid MRI-near-infrared (NIR) system has the potential to reduce false-positives and the number of follow-up invasive procedures in breast cancer diagnosis using complementary information from optical signatures. The computational toolbox proposed here will provide a powerful and efficient method for viable and more accurate three-dimensional imaging of large clinical subject populations in this framework. Overall, this will further advance the study of high-resolution optical signatures of normal and diseased breast tissue in-vivo and fluorescence imaging for studying biochemical and cellular mechanisms in-vivo.