ITR/CCR: Geometric Surface Processing Tools for Analysis of Biological Data

项目概述该项目的目标是开发一套新的计算工具，用于处理和分析从复杂生物数据集中提取的表面。其目的是缩小阻碍几何表面过滤工具在生物和医学应用中广泛使用的差距。这种差距的存在有两个原因。首先，这些用于过滤表面的复杂而强大的工具依赖于计算要求很高的数值算法。第二个原因是，生物应用不仅需要表面过滤，还需要一套用于高级处理的配套工具。因此，建议的工作将解决这两个问题，更好的过滤算法和更高级别的处理工具。这项工作的更广泛影响如下。大量的科学和医学调查依赖于通过3D成像研究受试者群体。这些研究产生了图像数据库，因此需要更好的工具从单个图像中提取有意义的信息，并量化几何特性，以便得出科学结论。以前的工作集中在提取嵌入在3D数据中的表面，但许多相关的生物学问题都与这些表面的几何特性有关。研究人员设想了一个强大的新工具箱，使科学家、工程师和临床医生能够处理3D表面数据。更具体地说，所提议工作的PI参与了几个具有此类需求的不同项目，因此所提议的工作对正在进行的具有直接应用的研究具有具体的含义。此外，提议的工作有一个教育元素，包括博士后候选人和研究生的多学科教育机会，以及一个项目，通过这个项目，我们将让本科生接触到3D图形硬件和软件的前沿研究。提出的工作建立在以前的结果，将图像处理的基本原理推广到表面。这项工作产生了从层析数据和变分图像滤波器（如高斯平滑、各向异性扩散和高升压滤波）的几何版本的表面重建技术。这些发展对于需要提取、分析和可视化表面的各种问题具有广泛的含义。为了实现这个项目的目标，研究人员将进行两个领域的研究。首先是发展更好的变分表面滤波算法。这部分工作的重点是速度和准确性。调查人员将研究更有效和准确的数字技术和表面表示，以及在商品图形卡上的实现。第二个研究领域是开发新的分析算法。这项工作将集中在表面分割、特征检测、形状量化和表面匹配上。所有这些工作都将与Insight Toolkit一起实施，Insight Toolkit是一个由nih赞助的用于多维图像处理的开源软件框架。

EIA-0313268Ross T. WhitakerUniversity of UtahProject SummaryThe goal of this project is to develop a new set of computational tools for processing and analyzing surfaces that are extracted from complex biological data sets. The purpose is closing the gap that prevents the widespread use of geometric surface-filtering tools in biological and medical applications. This gap exists for two reasons. The first is that these sophisticated and powerful tools for filtering surfaces rely on numerical algorithms that are computationally demanding. The second reason is that biological applications require not only surface filtering, but a set of complimentary tools for higher-level processing. Thus, the proposed work will address both issues, better filtering algorithms and higher-level processing tools. The broader impact of this work is as follows. Agro wing number of scientific and medical investigations rely on studying populations of subjects through 3D imaging. Such studies produce databases of images and therefore require better tools for extracting meaningful information from the individual images and quantifying geometric properties in order to make scientific conclusions. Previous work focused on extracting the surfaces that are embedded in this 3D data, but many of the relevant biological questions pertain to the geometric properties of those surfaces. The investigators envision a powerful new toolbox that will enable scientists, engineers, and clinicians to process 3D surface data. More specifically, the PI for the proposed work is engaged in several different projects that have such requirements, and thus the proposed work has concrete implications for ongoing research with direct applications. Furthermore, the proposed work has an educational element that includes multidisciplinary educational opportunities for postdoctoral candidates and graduate students and a program by which we will expose undergraduate women to leading-edge research in hardware and software for 3D graphics. The proposed work builds upon previous results that generalize the fundamentals of image processing to surfaces. That work produced technologies for surface reconstruction from tomographic data and geometric versions of variational image filters such as Gaussian smoothing, anisotropic diffusion, and high-boost filtering. These developments have broad implications for a variety of problems that entail extracting, analyzing, and visualizing surfaces. To satisfy the goals of this project, the investigators will pursue two areas of research. The first is the development of better algorithms for variational surface filtering. The focus of this part of the work is on speed and accuracy. The investigators will examine more efficient and accurate numerical techniques and surface representations, as well as implementations on commodity graphics cards.The second area of research is the development of new algorithms for analysis. This work will focus on surface segmentation, feature detection, shape quantification, and surface matching. All of this work will be implemented in conjunction with the Insight Toolkit, an open-source, NIH-sponsored software framework for multidimensional image processing.