Data Analysis Core

数据分析核心

基本信息

批准号：
10704487
负责人：
Jeffrey M Spraggins
金额：
$ 46.35万
依托单位：
VANDERBILT UNIVERSITY
依托单位国家：
美国
项目类别：
财政年份：
2020
资助国家：
美国
起止时间：
2020-09-30 至 2024-08-31
项目状态：
已结题

来源：
https://reporter.nih.gov/project-details/10704487
关键词：
3-Dimensional Address Age Algorithms Atlases Authorization documentation Automation Behavior Biological Assay Calibration Classification Computer software Confidence Intervals Cues Data Data Analyses Data Set Development Dimensions Disease Eye Funding Glean Goals High Performance Computing Histopathology Human Image Ions Link Maps Mathematics Measurement Measures Medical Imaging Methods Microscopy Mining Modality Modeling Molecular Multilingualism Multimodal Imaging Normal tissue morphology Normalcy Optical Coherence Tomography Organ Output Pancreas Patients Phase Play Reporting Resolution Sampling Scanning Source Specific qualifier value System Technology Tissue imaging Tissues Training Translating Variant Vendor Work analysis pipeline authority cell type computerized data processing data analysis pipeline data exchange data mining data quality data reduction data visualization deep learning file format imaging facilities imaging modality in vivo in vivo imaging inclusion criteria microscopic imaging multimodal data multimodality novel strategies open source parallelization reconstruction scaffold spatial integration temporal measurement tissue mapping whole slide imaging

项目摘要

PROJECT SUMMARY – Data Analysis Core. The VU-BIOMIC data analysis core (DAC) is tasked with automation of the reconstruction and subsequent analysis of the acquired multimodal eye and pancreas tissue imaging data. This is translated into four specific aims: (i) modality-specific data processing; (ii) data analysis pipeline development for 2-D and 3-D molecular tissue mapping; (iii) map construction for establishing 3-D molecular organization and function; and (iv) consortium coordination. In Aim 1, we will develop methods for preparing acquired measurement data for subsequent spatial integration, analysis, and content mining, and to remove any non-biological variation from the measurements prior to integration. In Aim 2, the DAC provides rapid cues for data quality assessment and ongoing multimodal analysis as new data is integrated into the atlases. Pre-analytically, we will develop data-derived sample inclusion criteria based on LC-MS/MS measurements, combined with gold standard histopathology, to capture what is “normal” tissue. To enable data mining of the massive 3-D multimodal spatially resolved datasets, accurate registration of multiple 2-D datasets into 3-D volumes will be essential. We will build a high-resolution mono-modal 3-D scaffold, using pre- measurement autofluorescence microscopy taken from every single tissue section. Furthermore, the 3-D data and analysis outputs, reconstructed from serial sections, will be spatially linked (by means of 3-D-to-3-D registration models) to the organ-specific in vivo and ex vivo 3-D scans to relate the acquired spectral data to more commonly encountered medical imaging modalities. Data-driven image fusion will enable the empirical discovery of potential correlative, anti-correlative, multivariate linear, and nonlinear relationships between observations in the different modalities, and also provide a framework for estimating to higher spatial resolutions as well as for out-of-sample prediction from one modality to another. The DAC will perform temporally resolved analysis of the data to find how molecular content changes with patient age. In Aim 3, the map construction phase, we will bring the third dimension to the varied data types that are measured and annotated. Data-driven image fusion will be used to advance the 3-D maps beyond what can be gleaned from one technology alone, including the application of IMS-AF-fusion-driven out-of-sample prediction. This will enable prediction of IMS observations at cutting depths where no IMS is measured. This will effectively provide predictive up-sampling of the 3-D tissue maps along the z-axis, building finer resolution 3-D volumes than would be possible with IMS alone. In Aim 4, we will develop specifications for the open file formats used in this work, multilingual parsers to ease access, and a URL-based Restful API to make (authorized) data exchange easy and accessible. We will work with the consortium to build common coordinate atlases based on in vivo images and continue the work of the currently funded project in specifying and developing easily disseminated file formats.

项目摘要 – 数据分析核心 VU-BIOMIC 数据分析核心 (DAC) 的任务是：对所获得的多模式眼睛和胰腺组织进行重建和后续分析的自动化这转化为四个具体目标：(i) 特定模式的数据处理；(ii) 数据分析； (iii) 建立 3-D 的地图构建分子组织和功能；以及 (iv) 联合体协调在目标 1 中，我们将开发方法。为后续的空间整合、分析和内容挖掘准备获取的测量数据，并在目标 2 中，DAC 提供了在积分之前消除测量结果中的任何非生物变异的功能。随着新数据集成到数据中，数据质量评估和持续多模式分析的快速提示在分析前，我们将制定基于 LC-MS/MS 的数据衍生样本纳入标准。测量与金标准组织病理学相结合，捕获“正常”组织以获取数据。挖掘海量 3D 多模态空间分辨数据集，准确配准多个 2D 数据集我们将使用预构建高分辨率单模态 3D 支架。测量自体荧光显微镜从每个组织切片此外，3-D 数据。从连续部分重建的分析输出将在空间上链接（通过 3-D 到 3-D 配准模型）到器官特异性体内和离体 3D 扫描，将采集到的光谱数据与更常见的医学成像模式将实现经验驱动的图像融合。发现之间潜在的相关、反相关、多元线性和非线性关系不同模式的观测，并提供估计更高空间分辨率的框架以及从一种模态到另一种模态的样本外预测，DAC 将执行时间解析。分析数据以了解分子含量如何随患者年龄变化。在目标 3 中，构建图谱。阶段，我们将把第三维度引入数据驱动的各种数据类型中。图像融合将用于改进 3D 地图，超出仅从一项技术所能收集到的范围，包括 IMS-AF 融合驱动的样本外预测的应用，这将使 IMS 的预测成为可能。在没有测量 IMS 的切割深度进行观察，这将有效地提供预测的上采样。沿 z 轴绘制 3D 组织图，构建比 IMS 更高分辨率的 3D 体积仅在目标 4 中，我们将为这项工作中使用的开放文件格式、多语言解析器制定规范。轻松访问，以及基于 URL 的 Restful API，使（授权）数据交换变得轻松且可访问。与该联盟合作，建立基于体内图像的共同坐标图集，并继续开展以下工作目前资助的项目是指定和开发易于传播的文件格式。