Automated Web-Based Behavioral Diagnostics of Cognitive Impairment

描述(由申请者提供)：阿尔茨海默氏症目前影响着大约530万美国人，这个数字预计在未来十年将显著增长。阿尔茨海默病研究的一个关键目标是改进目前的诊断方法，以便能够更快地识别患者，从而从现有的治疗方法中获得更大的优势。这个项目的主要目标是开发和验证一种自动化的、基于网络的方法，用于早期诊断认知能力下降和记忆丧失。我们将在现有研究的基础上，证明视觉配对比较任务(VPC)的表现与MCI诊断之间存在明确的联系，我们将通过开发一套软件工具和分析方法来扩大这一发现的影响，这些工具和分析方法将通过可广泛部署的基于Web的VPC任务来提高认知诊断的准确性和可及性。尽管VPC任务作为一种诊断辅助手段很有希望，但由于需要使用眼球跟踪器来精确监控受试者的眼球运动，临床应用受到严重限制。不幸的是，眼球追踪器价格昂贵，需要训练有素的人员，而且还不是很普遍。然而，我们的初步发现表明，在任何连接到互联网的计算机上，使用任何人都可以管理的vPC任务的修改版本，可以产生类似于眼动的图片检查行为。此外，来自计算机科学的强大的机器学习技术可以帮助准确分析和诊断由此产生的行为。这项工作的一个重要贡献将是有可能比现在更早地预测MCI患者即将到来的认知能力下降，也就是说，在神经系统受损较少、更有可能从治疗干预中受益的时候。为了支持这一目标，我们提出了三个目标：1)开发和研究适当的眼动特征表示和相应的基于机器学习的分类技术，以有效地识别患者状态；2)开发和验证基于网络的VPC任务版本；以及3)探索我们的任务作为普通老年人自动筛查工具的可行性。这些目标的成功完成有可能极大地改变目前临床转化研究的实践以及目前用于诊断认知缺陷的方法。这将使数以千计甚至可能数以百万计的患者和潜在的研究对象能够将这项测试作为常规检查的一部分，只需要一台连接互联网的电脑即可。

DESCRIPTION (provided by applicant): Alzheimer's disease affects an estimated 5.3 million Americans currently, and this number is expected to grow significantly in the coming decade. A critical goal of Alzheimer's disease research is to improve current methods of diagnosis so that patients can be identified sooner and, therefore, obtain greater advantage from available therapies. The major goal of this project is to develop and validate an automated, web-based method for early diagnosis of cognitive decline and memory loss. We will build on current research that has demonstrated a clear link between performance on the Visual Paired- Comparison Task (VPC) and diagnosis of MCI, and we will extend the impact of this finding by developing a suite of software tools and analysis methods that will improve the accuracy and extend the accessibility of cognitive diagnostics with a web-based VPC task that can be widely deployed. Although the VPC task is promising as a diagnostic aid, clinical application is severely limited by the need to use an eye tracker to precisely monitor subjects' eye movements. Unfortunately, eye trackers are expensive, require trained personnel, and are not widely available. However, our preliminary findings show that it is possible to produce picture examination behavior that is similar to eye-movements, using modified versions of the VPC task that could be administered by anyone, on any computer with an internet connection. In addition, powerful machine learning techniques from computer science can help accurately analyze and diagnose the resulting behavior. An important contribution from this work will be the possibility of predicting oncoming cognitive decline in MCI patients sooner than is now possible, that is, at a time when the nervous system is less compromised and more likely to benefit from therapeutic intervention. In support of this objective, we propose three aims: 1) Develop and investigate appropriate representation of eye movement characteristics and corresponding machine learning-based classification techniques for effective identification of the patient status, 2) Develop and validate a web-based version of the VPC task, and 3) explore the feasibility of our task as an automatic screening instrument for the general elderly population. Successful completion of these aims has the potential to dramatically alter the current practice of clinical translational research as well as the current methods used for diagnosing cognitive deficits. This would enable thousands and potentially millions of patients and potential research subjects to take the test as part of their routine checkup, requiring nothing more than a computer with an internet connection.