An AI-assisted screening platform within a multivariate framework for biomarkers of mild cognitive impairment due to Alzheimer's disease

项目摘要 越来越多的证据表明，每65秒，就有一个人患上老年痴呆症（AD）。 在美国，超过570万美国人患有这种疾病。阿尔茨海默氏症和其他痴呆症将花费 到2050年，国家将达到2770亿美元。主要问题是，许多认知障碍（CI）患者可能不知道 因为痴呆症的诊断和报告不足。缺乏低成本和非 侵入性筛查工具，以高准确性自动识别CI风险个体。因此，我们认为， 考虑到痴呆症流行的全球和社会影响，需要更好的战略来确定 有痴呆风险的患者。眼睛健康评估为我们的眼睛健康提供了一个独特的视角， 我们的身体例如，作为诊断模态的视网膜的视觉观察已经被广泛用于检测视网膜病变。 高血压，糖尿病，高胆固醇，甚至脑肿瘤，因为医生可以看到视神经， 大脑的一部分因此，眼睛测试也可能是检测CI的潜在解决方案。虽然早期 许多风险因素的表现（例如，糖尿病、高血压和心脏病）， 人类视网膜，他们可能混淆CI的第一个迹象。在医疗保健领域，数据量的复杂性和增长 为人工智能（AI）在医学中的应用引起了全球的极大兴趣。因此，我们认为， 我们的目标是通过人工智能提供一个实用的CI近期风险评估，通过识别和利用新的 多变量生物标志物（包括眼睛标志物）具有更好的区分能力。在本I期STTR中，iScreen 2 Prevent，LLC、迈阿密和iCareHub，LLC将开发一个基于人工智能的筛查平台， 早期发现AD引起的CI。我们的初步数据显示，多变量眼部生物标志物与 认知状态，并可用于区分轻度CI患者与认知健康受试者（年龄匹配 (55+岁），接受操作曲线下面积（AUROC）=0.90（SE=0.050），p<0.001）。然而，在这方面， 多变量生物标记物需要在筛选时组合以提高预测的准确性， 生物标志物方法学可以通过人工智能来推进。我们的目标是整合和优化我们的眼睛筛查 框架（iScreen 2 Predict™）整合到数字健康平台（iCAREHub）中， 在护理点的全面临床数据。我们还旨在开发一个基于AI的模型， 多变量标记物，并测试iScreen 2 Predict™软件区分轻度CI患者的能力， 到AD。该项目填补了AD诊断领域的关键技术空白。虽然筛选次数 使用单峰和昂贵的生物标志物的工具继续增长，这些工具不考虑多变量数据 在日常护理过程中以集体和自动化的方式产生。因此，其诊断潜力有限。 我们的软件开发作为医疗器械（SaMD），用于早期检测AD所致CI，考虑到 与认知状态相关的眼脑异常的多因素变量和相关生物标志物将允许 早期干预和促进更好地管理疾病的主要认知症状。

PROJECT SUMMARY Accumulating evidence indicates that every 65 seconds, someone develops Alzheimer's disease (AD) in the United States, and over 5.7 million Americans have the condition. Alzheimer's and other dementias will cost the nation $277 Billion by 2050. The major problem is that many people with cognitive impairment (CI) may not know they have it because dementia is underdiagnosed and underreported. There is a lack of low-cost and non- invasive screening instruments to identify individuals at risk for CI with high accuracy automatically. Therefore, considering the global and societal implications of the dementia epidemic, better strategies are needed to identify patients at risk for dementia. An eye health evaluation offers a unique perspective on the health of our eyes and our bodies. For example, visual observation of the retina as a diagnostic modality is already widely used to detect high blood pressure, diabetes, high cholesterol, and even brain tumors since a physician can see the optic nerve, which is part of the brain. Thus, an eye test may also be a potential solution to detect CI. While early manifestations of numerous risk factors (e.g., diabetes, hypertension, and heart disease) have been found in the human retina, they may confound the first signs of CI. In healthcare, the complexity and rise in data volume have contributed to the remarkable worldwide interest of Artificial Intelligence (AI) applications in medicine. Therefore, we aim to provide a practical near-term risk assessment of CI through AI, by identifying and utilizing novel multivariate biomarkers (including eye markers) with a better discrimination power. In this Phase I STTR, iScreen 2 Prevent, LLC, the University of Miami, and the iCareHub, LLC, will develop an AI-based screening platform for early detection of CI due to AD. Our preliminary data show that multivariate eye biomarkers are related to cognitive status and can be used to discriminate mild CI patients from cognitively healthy subjects (age-matched (55+ years old), area under the receiving operating curve (AUROC)=0.90 (SE=0.050), p<0.001). However, multivariate biomarkers need to be combined at the point of screening to enhance the accuracy of predictions, and biomarker methodologies could be advanced using AI. We aim to integrate and optimize our eye screening framework (iScreen 2 Predict™) into a digital health platform (iCAREHub) that collects personalized, comprehensive clinical data at the point of care. We also aim to develop an AI-based model with the integrated multivariate markers and test the iScreen 2 Predict™ software's ability to discriminate patients with mild CI due to AD. This project fills a critical technology gap in the field of AD diagnostics. While the number of screening tools using unimodal and expensive biomarkers continues to grow, these tools do not consider multivariate data generated during the routine care in a collective and automated way. Thus, their diagnostic potential is limited. The development of our software as a medical device (SaMD) for detecting CI due to AD earlier, considering multifactorial variables and relevant biomarkers of ocular-brain abnormalities related to cognitive status, will allow earlier intervention and facilitate better management of the disease's primary cognitive symptoms.