AitF: Collaborative Research: Topological Algorithms for 3D/4D Cardiac Images: Understanding Complex and Dynamic Structures

AitF：协作研究：3D/4D 心脏图像的拓扑算法：理解复杂和动态结构

• 批准号: 1733843
• 负责人: Dimitris Metaxas
• 金额: $ 26.6万
• 依托单位: Rutgers University New Brunswick
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-09-15 至 2020-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1733843&HistoricalAwards=false
• 关键词: AitF; Collaborative; Research; Topological; Algorithms

The interiors of ventricles of a human heart are spanned by a fine net of muscle fibers that are difficult to resolve, even in high resolution CT images. An accurate account of these structures, however, could improve diagnosis of cardiac disease, evaluation of cardiac function, assessment of stroke risk, and simulation of cardiac blood flow. Topology is the branch of abstract mathematics that deals with connections; this project uses the theory of persistent homology to identify crucial topological handles that can be useful for accurate reconstruction and analysis of the complex cardiac dynamics from these CT images. The outcome of the project will not only advance our understanding of cardiac function, but also generate novel computational topology methods that are more efficient and effective for practical applications. This project not only bridges the gap between the theory of computational topology and the practical problem of cardiac image analysis, but also trains the next generation of researchers and educators to do so by a carefully integrated education plan. The PIs will engage undergraduate students, high school students, women and other underrepresented students in their proposed research.The goal of this project is to develop a topological approach to unveil the intrinsic structures from complex and dynamic 3D/4D cardiac data, and furthermore, to provide principled tools to quantitatively analyze these structures. The PIs will create new computational topology methodologies and algorithms to extract rich information from the intrinsic structure of cardiac data. They will develop novel methodologies to extract localized topological features and to track them based on their spatial and temporal coherence. They also plan to design new algorithms to untangle ambiguous and uncertain situations for tracking structures through time sequence data. The resulting techniques and software will be validated on cardiac CT data to produce quantitative assessments of accuracy and to characterize the advantages and limitations of these approaches. Domain experts will validate the quality of the approaches via scientific hypotheses and data exploration. The methods to be developed are general and will impact other scientific fields where intrinsic complex and dynamic structures exist.

即使在高分辨率的CT图像中，也难以解决的肌肉纤维的细网跨越了人心脏的内部。 然而，对这些结构的准确说明可以改善心脏病的诊断，心脏功能的评估，中风风险评估以及对心脏血流的模拟。拓扑是处理连接的抽象数学分支。该项目使用持久同源性理论来识别至关重要的拓扑处理，这对于精确重建和分析了这些CT图像的复杂心脏动力学。该项目的结果不仅会提高我们对心脏功能的理解，而且还会产生新颖的计算拓扑方法，这些方法对实际应用更有效，有效。该项目不仅弥合了计算拓扑理论与心脏图像分析的实际问题之间的鸿沟，而且还培训了下一代研究人员和教育工作者，通过精心整合的教育计划来做到这一点。 PI将参与本科生，高中生，妇女和其他代表性不足的学生参与其拟议的研究。该项目的目的是开发一种拓扑方法，以揭示从复杂而动态的3D/4D心脏数据的内在结构，以及提供定量分析这些结构的原始工具。 PI将创建新的计算拓扑方法和算法，以从心脏数据的内在结构中提取丰富的信息。他们将开发出新的方法来提取局部拓扑特征，并根据它们的空间和时间连贯性跟踪它们。他们还计划设计新的算法，以使歧义和不确定的情况毫无障碍，以通过时间序列数据跟踪结构。最终的技术和软件将在心脏CT数据上进行验证，以产生准确性的定量评估，并表征这些方法的优势和局限性。领域专家将通过科学假设和数据探索来验证方法的质量。要开发的方法是一般的，将影响存在内在复杂和动态结构的其他科学领域。

项目成果

Deep attentive feature learning for histopathology image classificatication

用于组织病理学图像分类的深度注意力特征学习

• 发表时间: 2019
• 期刊: International Symposium on Biomedical Imaging (ISBI
• 作者: Wu, Pengxiang;Qu, Hui;Huang, Qiaoying;Chen, Chao;Metaxas, Dimitris
• 通讯作者: Metaxas, Dimitris

Point Cloud Processing via Recurrent Set Encoding

• DOI: 10.1609/aaai.v33i01.33015441
• 发表时间: 2019-07
• 作者: Pengxiang Wu;Chao Chen;Jingru Yi;Dimitris N. Metaxas

ASSD: Attentive Single Shot Multibox Detector

• DOI: 10.1016/j.cviu.2019.102827
• 发表时间: 2019-09
• 期刊: Comput. Vis. Image Underst.
• 作者: Jingru Yi;Pengxiang Wu;Dimitris N. Metaxas