AI-Based Identification of Rapid Glaucoma Progression to Guide Clinical Management and Accelerate Clinical Trials

项目摘要 青光眼是全球不可逆转失明的主要原因，预计将影响超过1.1亿人 在接下来的二十年里，世界各地的人们。这是一种退行性疾病，在两个方面都有很大的影响 患者的生活质量和医疗保健系统的成本。青光眼临床治疗的迫切需要 研究是准确地识别可能经历快速疾病进展(即，丧失)的患者的能力 视觉功能快速)。目前，估计一名患者的进展速度需要多次随访 在多年的过程中。这种识别进展的延迟会导致视力丧失并增加成本 关爱。它还影响了青光眼的临床试验，增加了研究新事物所需的时间和成本 治疗这种疾病的方法。此第一阶段STTR提案的目标是使用人工智能技术来 提高青光眼视网膜脱离进展诊断的准确性，缩短诊断时间。其主要结果是 我们的第一阶段计划将使一个基于人工智能的工具能够识别青光眼的快速进展，并将 立即准备用于1/2a期临床试验，因为不需要FDA批准。具体地说，我们将(1) 使用纵向光学相干层析成像(OCT)成像和视野(VF)测试数据集来训练人工智能 识别快速发展的青光眼患者的模型和(2)合并患者数据、临床测量、 并将治疗历史引入AI模型，进一步提高性能。将对人工智能模型进行训练和评估 结合了研究和现实世界的临床数据。这些数据集包括数以万计的图像， VF测试和临床记录收集自超过9000名青光眼患者的不同队列 十多年的历程。这些数据集为我们提供了一个独特的机会，不仅可以训练人工智能模型， 而且还将模型性能表征为患者人口统计、临床协变量、疾病 严重性和随访长度-提供关键背景，帮助临床医生更好地理解模型预测。 准确和早期的预测将对临床管理和临床试验都大有裨益。 青光眼。改善结果、降低患者护理和药物开发成本以及更快地开发 青光眼疗法使快速识别进展者的工具成为我们目标客户的一种有吸引力的产品， 制药公司和眼部护理专家。

Project Summary Glaucoma is the leading cause of irreversible blindness worldwide and is expected to affect more than 110 million people worldwide within the next two decades. It is a degenerative disease that has a large impact both in terms of patient quality of life and in costs to the healthcare system. A critical need in glaucoma clinical management and research is the ability to accurately identify patients likely to undergo rapid disease progression (i.e., lose visual function quickly). Currently, estimating the rate of progression for a patient requires several follow-up visits over the course of multiple years. This delay in identifying progression leads to lost vision and increases the cost of care. It also impacts clinical trials in glaucoma, increasing the time and cost needed to investigate novel therapies for the disease. The goal of this Phase I STTR proposal is to use artificial intelligence techniques to improve the accuracy and shorten the time for identifying raid progression in glaucoma. The primary outcome of our Phase I proposal will enable an AI-based tool to identify rapid glaucomatous progression and will be immediately ready for use in Phase 1/2a clinical trials as FDA approval is not required. Specifically, we will (1) use longitudinal optical coherence tomography (OCT) imaging and visual field (VF) testing dataset to train AI models to identify rapidly progressing glaucoma patients and (2) incorporate patient data, clinical measurements, and treatment history into the AI models to further improve performance. AI models will be trained and evaluated on a combination of research and real-world clinical data. These datasets include tens of thousands of images, VF tests, and clinical records collected from a diverse cohort of more than 9,000 glaucoma patients over the course of more than a decade. These datasets provide us with a unique opportunity to not only train AI models, but also to characterize model performance as a function of patient demographics, clinical covariates, disease severity, and follow-up length – providing critical context to help clinicians better understand model predictions. Accurate and early predictions would be of great benefit to both clinical management and clinical trials in glaucoma. Improved outcomes, reduced patient care and drug development costs, and faster development of glaucoma therapeutics make tools that quickly identify progressors an attractive product for our target customers, pharmaceutical companies and eye care specialists.