Neuroinformatics platform using machine learning and content-based image retrieval for neuroscience image data

该项目旨在开发NeuroManager™，这是一个用于高级解析的创新神经信息学平台， 存储、聚合、分析和共享复杂的神经科学图像数据。一项核心技术， NeuroManager中开发的将是使用语义检索的图像内容分析（ICARUS），一本小说， 智能神经影像管理系统，将使图像检索基于视觉外观或通过 语义概念ICARUS将使用机器学习应用于基于内容的图像检索（CBIR）， 和完善模型，总结微观和宏观图像外观，并自动分配 语义概念到神经图像。神经科学研究产生了广泛的，多方面的数据， 由于对原始数据的访问通常由源实验室维护，因此未得到充分利用。上 另一方面，在神经科学研究中共享复杂图像数据有许多优点，包括 有机会由其他科学家从另一个角度单独分析原始数据， 科学研究及其结果的可重复性。不幸的是，神经科学数据共享选项 满足了神经科学家的所有需求。为了解决这个问题，NeuroManager将包括 以下不同的，重要的创新：（i）处理二维（2D）和三维的多功能性 来自动物模型和人类的神经成像数据集;（ii）共享扩展的复杂数据集的功能 来自机构、实验室甚至公共领域的安全、隐私控制的范例;（iii）灵活性 在研究所的IT基础设施内或大多数基于云的虚拟化平台上实施NeuroManager， 环境，包括Azure、Google云服务和Amazon Web服务;（iv）最重要的是， ICARUS技术用于神经成像数据集的CBIR。NeuroManager对神经科学的益处 研究社区，药理学和生物技术研发，以及整个社会将促进 科学家和机构之间的合作，通过在一个 跨学科的氛围。这将为更好地理解神经病理学打开新的视野 与多种不同水平的人类神经精神和神经病症相关（即， 宏观上，微观上，亚细胞和功能上），最终导致改善的基础， 为复杂的脑部疾病开发新的治疗和预防策略。在第一阶段，我们将证明 通过开发原型软件，将在2D整个幻灯片上执行CBIR，验证了这种新技术的可行性 我们的合作者正在进行的研究项目中的整个小鼠大脑的冠状切片图像。工作 第二阶段的重点是开发商业软件产品，其中包括所有的创新 上面提到的。一种具有可比功能的竞争技术，可满足各种需求 用于现代神经科学研究，目前还没有商业或其他方式。

This project aims to develop NeuroManager™, an innovative neuroinformatics platform for advanced parsing, storing, aggregating, analyzing and sharing of complex neuroscience image data. A core technology that we will develop in NeuroManager will be Image Content Analysis for Retrieval Using Semantics (ICARUS), a novel, intelligent neuroimage curation system that will enable image retrieval based on visual appearance or by semantic concept. ICARUS will use machine learning applied to content-based image retrieval - (CBIR) to build and refine models that summarize microscopic and macroscopic image appearance and automatically assign semantic concepts to neuroimages. Neuroscience research generates extensive, multifaceted data that is considerably under-utilized because access to original raw data is typically maintained by the source lab. On the other hand, there are many advantages in sharing complex image data in neuroscience research, including the opportunity for separate analysis of raw data by other scientists from another perspective and improved reproducibility of scientific studies and their results. Unfortunately, none of the neuroscience data sharing options that exist today fulfill all the needs of neuroscientists. To solve this problem, NeuroManager will include the following distinct, significant innovations: (i) versatility for handling two-dimensional (2D) and three-dimensional neuroimaging data sets from animal models and humans; (ii) functionality to share complex datasets that extends secure, privacy-controlled paradigms from institutional, laboratory-based and even public domains; (iii) flexibility to implement NeuroManager within an institute’s IT infrastructure, or on most cloud-based virtualized environments including Azure, Google Cloud Services and Amazon Web Services; (iv) and most importantly, the ICARUS technology for CBIR in neuroimaging data sets. The benefit of NeuroManager for the neuroscience research community, pharmacological and biotechnological R&D, and society in general will be to foster collaboration between scientists and institutions, promoting innovation through combined expertise in an interdisciplinary atmosphere. This will open new horizons for better understanding the neuropathology associated with several human neuropsychiatric and neurological conditions at various levels (i.e., macroscopically, microscopically, subcellularly and functionally), ultimately leading to an improved basis for developing novel treatment and prevention strategies for complex brain diseases. In Phase I we will prove feasibility of this novel technology by developing prototype software that will perform CBIR on 2D whole slide images of coronal sections of entire mouse brains from ongoing research projects of our collaborators. Work in Phase II will focus on developing the commercial software product that will include all of the innovations mentioned above. A competing technology with comparable functionality, addressing the full breadth of needs for modern neuroscience research, is currently not available commercially or otherwise.