An Integrated Software Platform for Accelerating Image-Driven Ophthalmic Research and Driving New Insights and Endpoints to the Clinic

摘要 超过2000万患者患有老年性黄斑变性、糖尿病视网膜病变或青光眼。这些 退行性眼病经过几十年的发展，其患病率正在增加。视网膜成像技术，如 光学相干断层扫描和自适应光学眼底镜是调查和 眼科疾病的管理。从这些和其他成像方式获得的新的定量生物标记物对于 新兴眼科创新的临床翻译。然而，人工智能时代的生物标记物发展 智能需要大量带注释的图像和透明、可重现的过程，这为新的 对活体研究、数据共享和算法开发的管理需求。不幸的是， 当前的软件平台不能以满足特定要求的方式有效地集成这些数据 眼科， 我们在这项直接到第二阶段SBIR中的目标，与NIH数据科学战略计划的目标一致，是为了创建 用于收集、管理、分析和共享眼部图像和数据的综合平台(PaaS)。我们将延长 威斯康星医学院高级眼科成像计划(AoIP)开发的系统的能力 (MCW)，其中包括：(A)格子--降低成本、减少错误和改善通信的软件解决方案 (B)MOSAIC-一个图像处理平台和算法库，具有 传统和人工智能训练的算法；和(C)AoIP图像库-存储1578上的图像和数据的存储库 完全同意的人体研究对象。为了创建一体化平台，我们将解决四个目标：(A)扩展晶格 满足本地和多站点环境中学术和赞助研究的工作流程要求，包括 可扩展的直接集成与眼科检查相关的数据；(B)设计和实施混合(本地+云) 数据所有者和读者的存储库架构、数据模式、知识本体和查询架构；(C) 将晶格、存储库和马赛克集成并演示到连续眼科学工作流程中，以及(D)集成 并将晶格、存储库和马赛克演示为连续的眼科学工作流程。我们的集成翻译 成像平台将使眼科创新者能够更快地将挽救视力的见解和干预措施转化为临床， 更少的挫折感和更大的自信。 我们的建议填补了眼科数据科学和生物标记物开发领域的一个重要技术空白。而当 成像设备的数量和类型持续增长，开发和部署新的生物标记物和临床的工具 使用这些精致成像设备的终端并没有跟上步伐。有了这个计划，我们将使新一代 形象驱动的创新找到了走向临床的道路。

ABSTRACT More than 20 million patients suffer from age-related macular degeneration, diabetic retinopathy, or glaucoma. These degenerative eye diseases develop over decades, and their prevalence is increasing. Retinal imaging technologies such as optical coherence tomography and adaptive optics ophthalmoscopy are essential tools in the investigation and management of eye disease. New quantitative biomarkers derived from these and other imaging modalities are critical to the clinical translation of emerging ophthalmic innovations. However, biomarker development in the era of artificial intelligence requires large volumes of annotated images and transparent, reproducible processes, which places new demands on the management of living subjects research, data sharing, and algorithm development. Unfortunately, current software platforms are not effective in integrating these data in a manner that meets specific requirements in ophthalmology, Our goal in this Direct-to-Phase II SBIR, consistent with objectives of the NIH Strategic Plan for Data Science, is to create an integrated platform (PaaS) for the collection, curation, analysis, and sharing of ocular images and data. We will extend the capabilities of systems developed by the Advanced Ocular Imaging Program (AOIP), Medical College of Wisconsin (MCW), which include: (a) LATTICE - a software solution that reduces costs, reduces errors, and improves communications in the management of living-subjects research; (b) MOSAIC - an image processing platform and algorithm library with traditional and AI-trained algorithms; and (c) The AOIP Image Bank - a Repository that houses images and data on 1578 fully-consent human research subjects. To create the integrative platform, we will address four aims: (a) Extend LATTICE to meet the workflow requirements of academic and sponsored research in local and multisite environments, including the extensible direct integration of data relevant to ocular studies; (b) Design and implement a hybrid (local + cloud) REPOSITORY architecture, data schema, knowledge ontology, and query architecture for Owners and Readers of data.; (c) Integrate and demonstrate LATTICE, REPOSITORY and MOSAIC into a continuous ocular science workflow and (d) integrate and demonstrate Lattice, Repository and Mosaic into a continuous ocular science workflow. Our Integrated Translational Imaging platform will enable ophthalmic innovators to translate sight-saving insights and interventions to the clinic faster, with less frustration, and greater confidence. Our proposal fills an important technology gap in the field of ophthalmic data science and biomarker development. While the number and type of imaging devices continues to grow, the tools to develop and deploy new biomarkers and clinical endpoints using these exquisite imaging devices has not kept pace. With this program we will enable a new generation of image-driven innovation to find its way to the clinic.