Inter-modal Coupling Image Analytics

模态间耦合图像分析

• 批准号: 10530041
• 负责人: Theodore Satterthwaite
• 金额: $ 76.24万
• 依托单位: UNIVERSITY OF PENNSYLVANIA
• 依托单位国家: 美国
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-05-10 至 2027-06-30
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/10530041
• 关键词: Accounting; Address; Biological Markers; Brain; Brain imaging; Complex; Coupling; Data Scientist; Development; Diagnostic; Executive Dysfunction; Gene set enrichment analysis; Genomics; Goals; Human; Image; Left; Machine Learning; Maps; Measurement; Measures; Mental disorders; Methodology; Methods; Modality; Multimodal Imaging; Non-linear Models; Philadelphia; Psychopathology; Reproducibility; Research Personnel; Statistical Methods; Structure; Work; Youth; cohort; connectome; data resource; high dimensionality; imaging modality; imaging study; interest; outcome prediction; success; tool

PROJECT SUMMARY Almost all brain imaging studies now collect multiple imaging modalities, in an effort to derive measures of both structure and function from diverse imaging sequences. While quantitative data scientists have focused on machine learning approaches for predicting outcomes using multi-modal imaging, rigorous statistical methods for examining the relationship between imaging modalities have lagged behind. At present, the lack of statistical methodologies for assessing inter-modal coupling (IMCo) has left investigators with ad hoc solutions that lack statistical power and are prone to type I error, posing a threat to scientific rigor and reproducibility. In this application, we propose robust methods that leverage subject-specific measurements and use nonlinear modeling to address complex relationships in brain maps or networks, while accounting for important covariates (Aim 1). Furthermore, we will develop powerful approaches for assessing whether effects of interest (e.g., psychopathology, development) are enriched within brain networks (Aim 2). Assessment of this coupling between statistical associations and brain networks will capitalize upon tools from statistical genomics (e.g., gene set enrichment analysis) to provide principled methods for conducting enrichment analyses using high- dimensional, personalized brain networks. Finally, we will use these tools to delineate how trans-diagnostic executive dysfunction in youth with mental illness is related to abnormalities in structure-function coupling within brain networks (Aim 3). To do this, we will leverage three massive data resources: the Philadelphia Neurodevelopmental Cohort (PNC; n=1,601), the Healthy Brain Network (n=3,200), and the Human Connectome-Development (HCP-D; n=1,300) study Taken together, the proposed work builds upon the notable success in the first project period, promising to yield rigorous and generalizable methods for delineating the relationships between complementary measures of brain structure and function.

项目摘要 现在几乎所有的脑成像研究都收集多种成像方式，以获得两者的测量结果。 从不同的成像序列的结构和功能。虽然定量数据科学家专注于 使用多模态成像和严格统计方法预测结果的机器学习方法 检查成像方式之间的关系已经落后。目前，缺乏统计 评估模式间耦合（IMCo）的方法留给研究人员的是缺乏特定解决方案的解决方案， 统计能力，容易出现第一类错误，对科学的严谨性和可重复性构成威胁。在这 应用，我们提出了强大的方法，利用特定于主题的测量和使用非线性 建模以解决大脑地图或网络中的复杂关系，同时考虑重要的协变量 (Aim 1）。此外，我们将开发强大的方法来评估利益的影响（例如， 精神病理学、发展）在大脑网络中得到丰富（目标2）。对这种耦合的评估 统计关联和大脑网络之间的联系将利用来自统计基因组学的工具（例如， 基因集富集分析），以提供使用高- 三维的个性化大脑网络最后，我们将使用这些工具来描述如何跨诊断 精神疾病青年的执行功能障碍与其内部结构-功能耦合异常有关。 大脑网络（目标3）。为此，我们将利用三大数据资源： 神经发育队列（PNC; n= 1，601），健康大脑网络（n= 3，200）和人类 连接体发育（HCP-D; n= 1，300）研究总的来说，拟议的工作建立在值得注意的 在第一个项目期间取得了成功，有希望产生严格和普遍的方法， 大脑结构和功能的互补测量之间的关系。