Next generation axonal quantification and classification using AI

使用人工智能的下一代轴突量化和分类

基本信息

批准号：
10698843
负责人：
JACOB R GLASER
金额：
$ 87.3万
依托单位：
MICROBRIGHTFIELD, LLC
依托单位国家：
美国
项目类别：
财政年份：
2021
资助国家：
美国
起止时间：
2021-08-19 至 2026-06-30
项目状态：
未结题

来源：
https://reporter.nih.gov/project-details/10698843
关键词：
3-Dimensional Acute Address Animal Model Area Attention Autopsy Axon Biotechnology Brain Diseases Central Nervous System Classification Collaborations Communities Complex Computer Assisted Computer software Computers Data Detection Development Fiber Genes Goals High Performance Computing Human Image Individual Institution Investigation Knowledge Laboratories Learning Letters Manuals Measurement Methods National Institute of Mental Health Nature Neurodegenerative Disorders Neurodevelopmental Disorder Neurosciences Neurosciences Research Optics Pattern Performance Phase Physiological Presynaptic Terminals Production Research Research Personnel Running Societies Space Perception Specific qualifier value System Technology Testing Time Tissues Training Transgenic Animals Validation Work automated segmentation axon injury brain tissue cloud based cluster computing commercialization data format deep neural network density fighting image processing imaging modality improved innovation insight interest microscopic imaging nervous system disorder neural network neuropathology neuropsychiatric disorder new technology next generation novel open data pharmacologic prevent prototype reconstruction research and development software as a service terabyte tool treatment strategy two-dimensional usability web services

项目摘要

Abstract This Phase II project describes the commercial development of HyperAxon™, highly innovative software for performing automated segmentation, tracing, reconstruction and quantitative analysis of all axonal fibers (with and without signs of acute axonal injury) visible in two- and three-dimensional (2D and 3D) microscopy images of central nervous system (CNS) areas, even those with extremely high axonal fiber density. Accurate and rigorous analysis of all axonal fibers visible in 3D and 2D microscopy images of CNS tissue of non-transgenic and transgenic animal models as well as in human post mortem CNS tissue promises to enable researchers to gain novel insights into physiological neural network connectivity patterns as well as into the neuropathological underpinnings of alterations in connectivity associated with human neuropsychiatric and neurological disorders. However, this cannot be achieved with contemporary, computer-assisted tracing and reconstruction methods, which currently are the gold standard for investigating axonal fibers, because these methods primarily address tracing and reconstruction of only a limited number of individual axonal fibers. During Phase I we created HyperAxon prototype software by leveraging the original, lab-built technology Learning-based Tracing of Dense Axonal Fibers created at MIT Lincoln Laboratory (MIT LL) (Lexington, MA) and extending this technology with several new, specialized deep neural networks. Furthermore, we validated that our approach will be successful in research applications. All specific aims of Phase I were fully completed, demonstrating feasibility of successfully developing HyperAxon. The game-changing innovation in HyperAxon is the ability to automatically (i) segment, trace and reconstruct all axonal fibers visible in 3D and 2D microscopy images of CNS areas with high axonal fiber density, (ii) identify axonal branch points, (iii) resolve axonal fibers of passage in fiber tracts from those in axonal terminal fields, (iv) identify axonal fibers showing acute axonal injury and (v) precisely quantify alterations in number and density of axonal fibers in CNS tissue. For widespread dissemination of this important new technology we plan to commercialize the HyperAxon software at the end of Phase II as both a cloud-based “software as a service” running on Amazon Web Services (AWS) and traditional software application running on local institutional computers. We are convinced that HyperAxon will be impactful in the field of neuroscience research and will enable substantial advancements in research on alterations in CNS circuitry associated with neurodevelopmental, neuropsychiatric, neurodegenerative and neurological disorders. Ultimately, this will result in an improved basis for developing novel treatment strategies for a wide spectrum of complex brain diseases. In Phase I we demonstrated feasibility of this novel technology by developing prototype software; work in Phase II will focus on creating the full functionality of HyperAxon for commercial release. During Phase II we will perform extensive product validation and usability studies of HyperAxon in close collaboration with MIT LL and our academic collaboration partners. A competing technology is not available.

抽象的该第二阶段项目描述了 HyperAxon™ 的商业开发，这是一款高度创新的软件，用于对所有轴突纤维进行自动分割、追踪、重建和定量分析（并且没有急性轴突损伤的迹象）在二维和三维（2D 和 3D）显微镜图像中可见中枢神经系统 (CNS) 区域，甚至轴突纤维密度极高的区域。对非转基因 CNS 组织的 3D 和 2D 显微镜图像中可见的所有轴突纤维进行严格分析转基因动物模型以及人类死后中枢神经系统组织有望使研究人员能够获得对生理神经网络连接模式以及神经病理学的新见解与人类神经精神和神经系统疾病相关的连通性改变的基础。然而，现代计算机辅助追踪和重建方法无法实现这一点，目前是研究轴突纤维的黄金标准，因为这些方法主要解决在第一阶段，我们创建了仅有限数量的单个轴突纤维的追踪和重建。 HyperAxon 原型软件利用原始的实验室构建技术基于学习的密集追踪 Axonal Fibers 在麻省理工学院林肯实验室 (MIT LL)（马萨诸塞州列克星敦）创建，并将该技术扩展为此外，我们验证了我们的方法将会成功。第一阶段的所有具体目标均已全面完成，证明了其可行性。成功开发 HyperAxon 的颠覆性创新是能够自动执行任务。 (i) 分割、追踪和重建中枢神经系统区域 3D 和 2D 显微镜图像中可见的所有轴突纤维高轴突纤维密度，(ii) 识别轴突分支点，(iii) 解析纤维束中通道的轴突纤维从轴突末端区域中，（iv）识别显示急性轴突损伤的轴突纤维，以及（v）精确地改变中枢神经系统组织中轴突纤维的数量和密度的量化，以实现这一点的广泛传播。重要的新技术，我们计划在第二阶段结束时将 HyperAxon 软件商业化，作为在 Amazon Web Services (AWS) 和传统软件上运行的基于云的“软件即服务” 我们相信 HyperAxon 将在本地机构计算机上运行的应用程序产生影响。神经科学研究领域，将使中枢神经系统改变的研究取得实质性进展与神经发育、神经精神、神经退行性和神经系统疾病相关的电路。最终，这将为开发针对广泛疾病的新治疗策略奠定基础。在第一阶段，我们通过开发原型证明了这项新技术的可行性。软件；第二阶段的工作将集中于创建完整的 HyperAxon 以供商业发布。第二阶段，我们将密切合作对 HyperAxon 进行广泛的产品验证和可用性研究与 MIT LL 和我们的学术合作伙伴没有竞争的技术。